Isolation and Characterization of an Avian Hepatic Binding Protein Specific for N-Acetylglucosamine-terminated Glycoproteins*

An hepatic receptor which recognizes and binds specifically to serum glycoproteins bearing terminal, nonreducing N-acetylglucosamine residues has been purified to homoge-neity by affinity chromatography from chicken liver. The isolated binding protein has been characterized as a water-soluble glycoprotein in which sialic acid, galactose, mannose, and glucosamine comprise 8% of the total molecule. The binding reaction is a saturable process and is propor-tional to receptor concentration. Evidence has been adduced to indicate the presence of a single high affinity binding site with a dissociation constant of 1.4 x 10 9 M. A single subunit has been identified by polyacrylamide gel electrophoresis in sodium dodecyl sulfate with an estimated molecular weight of 26,000. The chemical and physical properties of the avian protein have been evaluated with respect to the analogous hepatic protein, of mammalian origin, which exhibits a binding specificity for galactose-terminal serum glycopro- teins. Removal of the terminal sialic acid from plasma proteins the mamma- lian the are the the


TOSHISUKE KAWASAKI~ AND GILBERT ASHWELL~
From the National Institute ofArthritis, Metabolism, and Digestive Diseases, National Institutes of Health, Bethesda, Maryland 20014 An hepatic receptor which recognizes and binds specifically to serum glycoproteins bearing terminal, nonreducing N-acetylglucosamine residues has been purified to homogeneity by affinity chromatography from chicken liver. The isolated binding protein has been characterized as a watersoluble glycoprotein in which sialic acid, galactose, mannose, and glucosamine comprise 8% of the total molecule. The binding reaction is a saturable process and is proportional to receptor concentration.
Evidence has been adduced to indicate the presence of a single high affinity binding site with a dissociation constant of 1.4 x 10 9 M. A single subunit has been identified by polyacrylamide gel electrophoresis in sodium dodecyl sulfate with an estimated molecular weight of 26,000. The chemical and physical properties of the avian protein have been evaluated with respect to the analogous hepatic protein, of mammalian origin, which exhibits a binding specificity for galactose-terminal serum glycoproteins. Removal of the terminal sialic acid from plasma proteins results in the formation of asialoglycoproteins bearing galactosyl residues as the newly formed, terminal sugar.
In mammalian species, the latter are recognized by an hepatic receptor whereby the entire protein is removed from the circulation, transported into the hepatocyte, and catabolized in the lysosomes (1   indicating the isolated binding protein to be essentially free from major contamination with inert protein.

Polyacrylamide
Gel Electrophoresis -The marked effect of detergent on the physical state of the isolated protein is illustrated in Fig. 2. Polyacrylamide gel electrophoresis of the binding protein, in a solution equilibrated with 0.1% Triton X-100, resulted in the formation of a single band (Fig. ZA); electrophoresis of the same protein, in the absence of added detergent, gave rise to the appearance of multiple components (Fig. 2B). The latter phenomenon, observed previously with the rabbit liver binding protein (31, was interpreted as reflecting a tendency towards self-association in aqueous solution which is common to many proteins of 'membranous oriein.

B
FIG. 2. Polyacrylamide gel electrophoresis of the binding protein and its subunit. A, the binding protein (10 pg) was dialyzed against a buffer containing 0.1% Triton X-100, 10 rn~ Tris/chloride, pH 7.8, and 10% glycerol at 4" for 2 days and applied to a 5% acrylamide gel. Electrophoresis was carried out using a multiphasic buffer system "System A," pH 9.45, for 2'12 hat 4". Protein bands were stained with Coomassie brilliant blue G-250 in 12.5% trichloroacetic acid; see "Materials and Methods" for details. B, an aqueous solution of the binding protein (15 pg) was applied to a 6.5% acrylamide gel. Electrophoresis and staining were carried out as described in A. The arrow denotes the position of the protein band seen when Tritoncontaining solutions of the binding protein were run at this acrylamide concentration. C, an aqueous solution of the binding protein (15 yg) was incubated in 1.0% sodium dodecyl sulfate, 10 rnM sodium phosphate, pH 7.0, for 3 days at 48". The solution was then made 10% in /3-mercaptoethanol, 0.5 mM EDTA, and 5 M in urea. Incubation was continued for an additional 3 h at 48" before applying the sample to a 10% acrylamide gel containing 0.1% sodium dodecyl sulfate. Electrophoresis was carried out at room temperature for 6 h according to the method of Weber and Osborn. The gels were soaked overnight in 10% acetic acid in 50% ethanol, prior to staining with Coomassie brilliant blue G-250 in 12.5% trichloroacetic acid.
Further characterization of the individual aggregates was not pursued. However, an estimate of the molecular weight of the single band observed in Fig. 2A was sought by means of a Ferguson plot (14). To this end, a series of five separate electrophoretic analyses of the detergent-equilibrated binding protein were carried out in which the total acrylamide concentration was varied from 4 to 6.5%. The resulting plot of gel concentration versus log RF yielded a straight line with a correlation coefficient of 0.999. The slope of this line, the retardation coefficient (K,J, was related to the molecular weight as shown in Fig. 3. The linear regression line of (KR)'i2 versus (molecular weighQ1'" was calculated by the method of least squares, based on a series of proteins with known molecular weights in the range of 70,000 and 500,000. The average molecular weight of the binding protein estimated from two separate experiments was 210,000. Subsequent investigation of subunit composition revealed the chicken binding protein to be resistant to complete dissociation under the standard conditions of denaturation described by Weber and Osborn (12). However, incubation for 3 days at 48" in 1% sodium dodecyl sulfate, followed by an additional 3-h exposure to 10% P-mercaptoethanol in 5 M urea gave rise to the single subunit seen in Fig. 2C. Alternatively, brief exposure of the protein to alkaline pH, as described under "Materials and Methods," resulted in the formation of a single band of identical mobility. This electrophoretic pattern was highly reproducible from experiment to experiment although, occasionally, an individual preparation of the binding protein contained a minor constituent migrating slightly ahead of the main band. The significance of this material, if any, is unknown. Evidence indicative of the presence of carbohydrates associated with the single subunit (Fig. 2C) was obtained by staining the gel with the periodate-Schiff reagent. The sole reactive area (not shown) coincided with that of the protein band.
In an attempt to estimate the molecular weight of the subunit, the binding protein was dissociated in sodium dodecyl sulfate as described above and subjected to electrophoresis at five different acrylamide concentrations ranging from 6 to 14%. In the resulting Ferguson plots, the subunit gave a linear regression line with a correlation coefficient of 0.999 and a free mobility (intercept of 0% acrylamide) closely comparable to that of a number of protein standards. Utilizing the linear regression of KR and molecular weight, calculated on a series of proteins with known molecular weights in the range of I  I  I  I  I  I  I   I  I  1  I  I  I  I  30   40  50  so  70   However, the association rate conterminated) or ahexosamino-orosomucoid (mannose-termi-stant for the reaction (K,) could not be assessed with certainty nated) as shown in Fig. 5A. This behavior contrasts strongly due to the lack of a reliable method for stopping the reaction with that of the rabbit liver binding protein which exhibited a instantaneously without dissociating the complex already clear nreference for asialo-orosomucoid and a low but measur-formed. able affinity for the corresponding glucosamine-or mannoseterminated derivatives (Fig. 5B). Utilizing the inhibition assay described previously (31, these studies were extended to a variety of analogously modified proteins (Table III) In contrast, dissociation was a relatively slow process and the reversibility of binding was examined as shown in Fig. 6. Following incubation of the binding protein with 12"1-agalactoorosomucoid, a loo-fold excess of nonradioactive ligand was added; aliquots were removed at various times and assayed for bound radioactivity.
In order to minimize any continuing dis- sociation taking place after precipitation with ammonium sulfate, the precipitate was immediately chilled in dry ice/acetone and processed quickly. In the presence of excess ligand, the reaction was shown to be of first order and the dissociation rate constant (K,) was calculated to be 1.3 x lo-" s-l.
As shown in Fig. 7, binding of 'Y-agalacto-orosomucoid to the purified avian membrane protein is a saturable process and, at appropriate ligand concentrations, binding is proportional to the protein concentration (Fig. 8). A Scatchard plot (26) of the binding data indicated the presence of a single, high affinity binding site with a dissociation constant of 1.4 x lo-" M, plotted as the average of two separate experiments yielding values of 1.34 and 1.43 x 10m9 M, respectively ( Fig. 9). At saturation, the maximum capacity of the binding protein was 3.9 pmol of '""I-agalacto-orosomucoid per pg of protein. On the assumption that the molecular weight of the binding protein is 210,000, it can be estimated that 0.8 mol of ligand was bound per mol of binding protein.
Stability of the Ligand .Binding Protein Complex -After an initial incubation carried out at pH 7.8, as described under "Materials and Methods," the pH of the mixture was adjusted to the values shown in Fig. 10  was observed up to and including pH 10.5. When the above procedure was modified to examine the pH dependency of the initial binding reaction, as opposed to the stability of the preformed complex, an entirely similar result was obtained. The stability of the complex formed at pH 7.8 was shown to be markedly temperature-dependent. At 0" the reaction product was stable for at least 2 h whereas after 1 h of incubation at 25, 37, and 45", respectively, 16, 37, and 58%  Control experiments, in the absence of the above reagents, exhibited complete retention of binding activity after a 30-min incubation at room temperature.

Effect of Glycosidases
-In an earlier study on the purified mammalian binding protein isolated from rabbit liver, neuraminidase effectively destroyed the ability of that protein to bind asialo-orosomucoid (3). In marked contrast, neuraminidase was without significant effect upon the ability of the avian protein to bind agalacto-orosomucoid (Fig. 11). However, in the presence of 1 milliunit of neuraminidase, P-galactosidase, even at levels as low as 0.2 milliunit, completely abolished the binding capacity of the chicken liver protein (curue A). That the latter effect was due to loss of binding capacity rather than to dissociation of ligand from the preformed complex is illustrated in curue B of Fig. 11. In this case, neuraminidase and /3-galactosidase were added subsequent to the incubation of binding protein and ligand.

Amino
Acid Composition -The amino acid composition of the avian binding protein is provided in Table IV. The number of amino acid residues per mol of subunit was based on a minimal molecular weight of the polypeptide portion of 5,760, with histidine set at unity, and the estimated molecular weight of the subunit (26,000)  -Initial evidence indicating the presence of carbohydrate in the purified binding protein was obtained by a positive periodic acid-Schiff stain on polyacrylamide gel electrophoresis which coincided with the single protein band. This finding was confirmed and extended by the data in Table V. The neutral sugars were determined by automated column chromatography, gas-liquid chromatography, and enzymatic assay after acid hydrolysis as described under "Materials and Methods." Good agreement was ob-  Sialic acid (2) Galactose (2) (2) Mannose (2) (1) Glucosamine tained by all three methods with the exception of the single low value for galactose determined by gas-liquid chromatography. No detectable amount of galactosamine, fucose, or glucose was observed by any of the above techniques. From the sum of the recorded values, the carbohydrate content of the purified binding protein was estimated to be 8%. The phosphorous content of the ashed protein was determined to be 0.06%, a value corresponding to less than 1 mol/mol of subunit. The finding was interpreted as indicating the absence of significant amounts of phospholipid or phosphorylated sugars in the final preparation.
Erythrocyte Agglutination -The ability of the rabbit binding protein to agglutinate human and rabbit red blood cells was reported earlier by Stockert et al. (27). Employing the incubation conditions used by these investigators, 7 to 10 pg of the avian binding protein produced no visible macroscopic agglutination of chicken, goat, sheep or human (A, B, 0) cells after 30 min at room temperature.
With rabbit and rat erythrocytes a minimal degree of agglutination was observed which was not further quantitated.

DISCUSSION
The avian protein described here shares many properties in common with those of the mammalian protein.
In both systems, the major locus of binding activity is restricted to the liver, from which it can be isolated, purified, and assayed by methods closely analogous to those described previously for the rabbit protein (3,4). In both cases, calcium is required for a binding reaction specifically directed toward the terminal, nonreducing carbohydrate residues of circulating proteins. Upon purification, both preparations proved to be glycoproteins containing the same set of carbohydrate constituents, including sialic acid, galactose, mannose, and glucosamine. In aqueous solution, the two proteins were recovered in an aggregated state which reversibly converted to a single molecular weight component by the addition of detergent. Finally, the binding of both proteins was abolished by the action of specific glycosidases.
However, within the framework of the above generalizations, specific properties are clearly distinguishable.
As shown in Table III and Fig. 5, the avian protein, in contrast to the mammalian, exhibits only minimal binding activity for asialoglycoproteins and interacts strongly with agalactoglycoproteins. A second major point of differentiation is the ready reversibility of the avian protein. ligand complex. From the data plotted in Fig. 6, a dissociation rate constant (K-J was calculated to be 1.3 x 10m:'sml. Again, in contrast to the kinetic properties of the mammalian protein, the attainment of equilibrium conditions in the avian system permitted the demonstration of a single, high affinity binding site with a K,,,,, of 1.4 x lo-!' M as determined by a Scatchard plot (Fig. 9).
Although both receptors were shown to be glycoproteins of similar carbohydrate composition, their response to specific glycosidases was characteristically different. Whereas the binding activity of the rabbit protein was destroyed by exposure to neuraminidase (31, the chicken protein was unaffected (Fig. 11). The activity of the latter preparation was abolished only after P-galactosidase was added to the incubation mixture.
Subunit structure provides another parameter of differentiation. The rabbit protein was shown to consist of two different subunits of 48,000 and 40,000 daltons, respectively (4). From the data reported here, the avian protein contains a single subunit with an estimated molecular weight of 26,000 (Fig. 4); the latter value was subsequently approximated as the sum of the amino acid (23,200) and carbohydrate (2000) residues (Tables IV and V). Unfortunately, the estimated minimum molecular weight of the avian protein cannot be assigned with confidence due to the obligate presence of detergent during the electrophoretic run (Fig. 3). However, if it can be assumed that the value of 210,000 represents a reasonable approximation, it can be calculated that the intact protein is composed of eight identical subunits and that 0.8 mol of ligand are bound per mol of binding protein.
Recent studies in vivo have demonstrated the hepatic uptake of agalactoglycoproteins in rats by a receptor system presumably different from that described earlier for asialoglycoproteins (28, 29). Whether, or to what extent, these observations reflect the presence of a specific mammalian receptor analogous to that described here is currently unknown. However, it should be noted that the purified mammalian binding protein exhibited a small but significant affinity for agalactoorosomucoid; in contrast, the avian protein was totally devoid of binding activity for asialo-orosomucoid (Fig. 5).