Use of Glycosyltransferases to Restore Pertussis Toxin Receptor Activity to Asialoagalactofetuin*

Fetuin derivatives with enzymatically altered oligo- saccharide units were tested for their ability to inhibit pertussis toxin-mediated agglutination of goose eryth- rocytes and the binding of 1261-labeled fetuin to pertussis toxin-coated polystyrene tubes. Fetuin oligosaccha- rides were sequentially degraded by treatment with: neuraminidase (asialofetuin) followed by 8-galactosid-ase (asialoagalactofetuin) and, lastly, with 8-N-acetyl- hexosaminidase (asialoagalacto-a[N-acetylglucosa-minolfetuin). Asialofetuin retained only 19 and 53% of the inhibitory activity of native fetuin in the hemagglutination and lZ6I-fetuin binding assays, respec- tively. Asialoagalactofetuin showed no further reduction of inhibition in the hemagglutination system and, instead, resulted in partial recovery of inhibition in the ‘261-fetuin-pertussis toxin binding assay. Asialoa-galacto-a[N-acetylhexosamino]fetuin showed a fur- ther decrease in ability to inhibit pertussis toxin binding in both assays. The inhibitory activity of asialoagalactofetuin could be restored to that of native fetuin by adding back D-galaCtOSe with UDP-Ga1:D-glucosyl-1,4-~-galactosyltransferase, The suggested that a pertussis may be the of in the of fetuin derivatives was 100 times in molar excess of that of the '"I-fetuin. The binding reaction was allowed to proceed for min at room temperature. labeled fetuin solutions were then discarded, and tubes were washed four with 0.5 ml of 0.1% BSA-buffer A. The of bound radioactivity was determined in an LKB Rack Gamma 1270 y coun- ter. The amount of lZ6I-fetuin bound to BSA-saturated polystyrene tubes which had not been precoated with pertussis toxin was 4% of the amount bound to pertussis toxin-coated, BSA-saturated tubes. In addition to unlabeled fetuin, excess unlabeled pertussis toxin was also able to compete for '=I-fetuin binding to pertussis toxin-coated

Pertussis toxin is one of the virulence factors produced by Bordetella pertussis, the etiological agent of whooping cough (Muiioz, 1985). It is the component best correlated with protection in the current whole cell pertussis vaccine (Muiioz et al., 1981;Sat0 and Sato, 1984), and most likely responsible for the neurological sequelae sometimes associated with the disease and in a small number of those vaccinated (Miller et al., 1981;Steinman et al., 1985). Consequently, there is considerable interest in elucidating structure-function relationships in pertussis toxin with a view to understanding its role in the disease process and for producing a safer acellular vaccine preparation.
Pertussis toxin provides another example of an A-B class of toxin (Gill, 1978). It is a heterohexameric protein consisting of five subunits, designated SI to S6 (Peppler et al., 1985; * This work was supported by grants from the Alberta Heritage Foundation for Medical Research and the Medical Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ TO whom correspondence should be addressed. Tamura et al., 1982). The largest of the subunits, SI, (the A subunit) is responsible for ADP-ribosylating the a components of a family of homologous, guanine nucleotide-dependent regulatory complexes (designated G or N proteins) in eukaryotic cells (Katada et al., 1983;Manning et al., 1984;Milligan et al., 1986;Moss and Vaughan, 1985;Sekura et al., 1983;Sternweis and Robishaw, 1984;Van Dop et al., 1984). The other four subunits of pertussis toxin probably form a pentameric base (B) structure composed of two dimers (S& (dimer 1) and S3S4 (dimer 2)) which are connected to each other by the Sg subunit (Tamura et al., 1982). As with other A-B toxins, the function of the base structure is in binding to host cell receptors and, possibly providing a means for the SI subunit to penetrate the cytoplasmic membrane (Nogimori et al., 1986;Tamura et al., 1982Tamura et al., , 1983.
Although the composition of pertussis toxin is well defined, the role of its structure in pathophysiology is incompletely understood. While most of pertussis toxin's effects may be related to the ADP-ribosyltransferase activity of the SI subunit (Nogimori et al., 1986), pertussis toxin's mitogenic activity, which may be responsible for lymphocytosis, is apparently the result of the interaction of the basal components with host cell receptors (Nogimori et al., 1984a(Nogimori et al., , 1984b. Accordingly, it is important to identify and characterize receptors for pertussis toxin in order to fully understand its activity. In addition to receptors in host tissues, pertussis toxin also interacts with receptors on chicken, horse, and particularly goose erythrocytes (Arai and Sato, 1976;Sekura and Zhang, 1985). Pertussis toxin also binds to the oligosaccharide portions of serum glycoproteins such as haptoglobin and fetuin (Irons and MacLennan, 1979;Sekura and Zhang, 1985). Although the molecular details of pertussis toxin's interaction with fetuin's carbohydrate groups have already been partially explored (Sekura and Zhang, 1985), the structure of fetuin's oligosaccharide units has subsequently been reevaluated and more precisely defined (Edge and Spiro, 1987;Takasaki and Kobata, 1986;Townsend et al., 1986). Moreover, the relationship between fetuin and pertussis toxin erythrocyte receptors or pertussis toxin receptors in host tissues has not been determined. Therefore, we reasoned it would be appropriate to reexamine the basis for the interaction of pertussis toxin with fetuin to gain further insight into the lectin-like properties of this toxin. In addition to providing greater insight into pertussis toxin's activity, receptor studies will prove useful in the identification of important targets for immunological neutralization of toxin activity and affinity matrices suitable for obtaining sufficient quantities of vaccine grade pertussis toxin.
One approach to elucidating the structural requirements for lectin recognition of oligosaccharide receptors is to sequentially remove monosaccharides from the receptor's nonreducing ends with exoglycosidase enzymes and determine the 8677 Sugar Groups Involved in Pertusis Toxin Binding to Fetuin effect this has on binding activity. However, the information gained from this approach is limited because: ( a ) receptor recognition may be dependent on the sugar's glycosidic linkage (Kronis and Carver, 1985;Wright, 1984); and ( b ) the specificity of exoglycosidase enzymes is not restricted to particular oligosaccharide sequences, a number of which are present in fetuin (Edge and Spiro, 1987;Spiro and Bhoyroo, 1974;Takasaki and Kobata, 1986;Townsend et al., 1986). However, the missing information can be obtained by using glycosyltransferases to add sugars back, in defined linkages, to known acceptor groups on exoglycosidase-trimmed oligosaccharides, thereby restoring receptor activity (Carroll et al., 1981;Rogers and Paulson, 1983). This approach not only allows the investigation of receptor structure and function at the primary chemical level, it has the advantage of maintaining the natural molecular environment for the restructured oligosaccharide units.
Buffers-Buffer A 0.  for procedures requiring extended periods of incubation at 37 'C.
Treatment of Fetuin with Exoglycosidases-One hundred mg of fetuin was dissolved in 4 ml of buffer B, and the solution was transferred to a 12 x 75-mm capped polystyrene culture tube. A portion (1 ml) of this solution was set aside in a separate culture tube for the control. Forty milliunits of neuraminidase (250 pg) was then added to the remaining 3-ml portion of fetuin, and the two culture tubes were incubated for 18 h at 37 "C while slowly turning (5 rpm) in an end-over-end rotator. After an additional 40 milliunits of fresh neuraminidase was added, the samples were incubated for an additional 18 h at 37 "C to complete the desialation reaction. Next, the samples were diluted 20 times with buffer B, and 0.4 units (250 pg) of bovine testes @-galactosidase was added to the mixture. The samples were then concentrated 20-fold using an Amicon Model 8010 ultrafiltration apparatus fitted with a 25-mm diameter YM-10 filter (molecular weight 10,000 cutoff) and incubated for 72 h at 37 "C to remove galactose residues. The samples were again diluted 20 times with buffer B, and 1 unit (130 pg) of jack bean 8-N-acetylhexosaminidase was added and the mixture concentrated and incubated at 37 'C for 72 h to remove N-acetylhexosamine groups. Next, the samples were diluted at least 20-fold in buffer A and concentrated in the Amicon ultrafilter. This procedure was repeated three times to remove any remaining sugars which had been released from protein by the enzyme treatments. In addition, at each stage of the procedure, a 100pl sample was removed and set aside for sugar analyses, protein concentration determination, and the binding-inhibition assays. The derivatives were stored at -20 "C until use.
Treatment of Asialoagaluctofetuin with Glycosyltransferases-A portion (6 mg) of the asialoagalactofetuin was diluted to 20 ml in buffer C and concentrated to 0.5 ml in the Amicon ultrafiltration unit as described above. This procedure was repeated. Next, 1.25 units of galactosyltransferase (313 pg) and 2 mg UDP-galactose (2 mol of UDP-Gai/mol of GlcNAc acceptor groups) were added, and the mixture was incubated for 24 h at 37 "C. A portion of the sample was removed for the hemagglutination inhibition assays and sugar analysis, and the remainder was diluted 20 times in buffer D. This was concentrated, and 10 milliunits of the sialyltransferase (1.25 pg) and 2 mg of CMP-N-acetylneuraminic acid (approximately 2 mol of CMP-NeuAc/mol of Gal acceptor groups) were added. After incubation for 24 h at 37 'C the reaction mixture was diluted 20 times in buffer A and concentrated to 0.5 ml. This final step was performed three times to ensure that all of the free monosaccharides were removed from the fetuin solution. Each of the aglycofetuin derivatives were diluted 25 times in buffer A to a protein concentration of roughly 0.5 mg/ml. The protein concentrations of the inhibitors was determined by the Lowry procedure (Lowry et al., 1951) using native fetuin for the standard. The protein concentrations were confirmed spectrophotometrically using an extinction coefficient of 4.1 for a 1% solution at X 278 nm (Spiro, 1960).
Sugar Analysis by Gas-Liquid Chromatography-Sugars were analyzed by gas-liquid chromatography (GLC) of the trifluoroacetate derivatives of the 0-methyl glycosides essentially as described previously (Zanetta et al., 1972). The analyses were performed on a Varian Vista 6000 (Varian Associates, Sunnyvale, CA) gas chromatograph equipped with a flame-ionization detector and a fused silica capillary column (30 m X 0.259 mm inner diameter) with a bonded and crosslinked nonpolar liquid phase 0.25 pm thick (a DB-5 column from J & W Scientific Co., Rancho Cordova, CA). Injection was performed in the splitless mode (0.5 min) followed by a 1001 split flow. The carrier gas (helium) had a linear velocity of 42 cm/s at the starting temperature of 90 "C. The column temperature was maintained at 90 "C for 4 min after the sample was injected and then programmed to increase at a rate of 8 'C/min to 270 "C. The injector and detector temperatures were 260 and 272 "C, respectively. Data processing was performed with a Varian 401 chromatograph data system. The relative amount of each sugar in the exoglycosidase-treated fetuin derivatives was obtained by comparing the area of each sugar peak relative to the area of the mannose peak in each sample. The reduction in sugar content in the aglycofetuin derivatives was expressed as the percentage of the mannose:sugar ratios obtained for each derivative as compared with those obtained for native fetuin. Sugars were identified by comparison with authentic samples obtained from Supelco.
To prepare the samples for analysis by GLC, fetuin and the aglycofetuin samples were diluted in 20 volumes of distilled water and concentrated in an Amicon Model 3 micro-ultrafiltration apparatus as described earlier. This procedure was repeated three times to remove most of the buffer salts and low molecular weight carbohydrate material (free sugars). Samples (200-500 pg) were then transferred to 1.5-ml amber, screw-cap septum vials (Pierce Chemical Go.) which had been rinsed with acetonitrile and dried with NZ gas prior to use. Next, the samples were dried in a Savant Spin-Vac concentrator (Model SVC-100H, Savant Instruments Inc., Hicksville, NY) and dissolved in 1.0 ml of 2.0 N anhydrous methanolic HC1 (3 N methanolic HCl from Supelco, diluted 2 1 in anhydrous methanol prior to use). The vials were purged with dry N z gas, tightly sealed with Teflon-lined screw caps, and placed in a heating block at 85 "C. After the initial warm-up period (-5 rnin), the vials were shaken vigorously with a Vortex mixer to ensure complete dissolution of the sample, and the caps were tightened further. Methanolysis was allowed to continue for 18 h at 85 'C. The caps were then removed from the cooled sample vials and the methanolic HCl was evaporated at 45 'C in a gentle stream of dry Nz gas. One hundred pl of trifluoroacetic anhydride:acetonitrile (13, Aldrich, and Caledon Labs., Georgetown, Ontario, Canada, respectively) was added, and the vials were sealed and incubated at 65 "C for 1 h. The samples were cooled, thoroughly dried in a stream of dry nitrogen, dissolved in 100 pl of anhydrous acetonitrile, and 1-2 p1 was injected into the chromato-

Goose H e~~g l~i~~i o n I n h i b i~n
Assay-Pertussis toxin was suspended to a concentration of 1 pg/ml (by weight protein) in buffer A and sonicated for 20 s in a Branson Model B-220 ultrasonic water bath, We found it necessary to sonicate diluted pertussis toxin solutions before each experiment to obtain consistent hemagglutination titer end-points. Next, serial 2-fold dilutions of the inhibitor solutions were prepared in buffer A, and 50 pl of each dilution was added to the round bottom wells in 96-well microtiter plates. Fifty pl of sonicated pertussis toxin solution was then mixed with the inhibitors, and the plates were incubated for 30-45 min at room temperature to allow binding to occur. The goose erythrocytes were suspended at a concentration of 1-4 X lo7 cells/ml in buffer A, and 50 pl was added to each well after the initial incubation period. The microtiter plates were then shaken and incubated at room temperature and observed for hemagglutination after 1-2 h. The minimum inhibitory concentration for fetuin and each of the aglycofetuin derivatives was then calculated. The activities of the derivatives were expressed relative to fetuin by dividing the minimum inhibitory concentration of fetuin by the minimum inhibitory concentration of each derivative.
'"I-Fetuin Polystyrene Tube Binding Inhibition Assays-Iodinated fetuin was prepared by the conventional Iodo-Gen procedure as described previously (Armstrong and Peppler, 1987) except that 40 pg of fetuin was exposed to Iodo-Gen for 5 min. The polystyrene tube binding assay was performed as described previously (Sekura and Zhang, 1985) but with some modifications. The polystyrene tubes were coated with 100 p1 of pertussis toxin (5 pg/ml in buffer E) for 18 h at 4 ' C. The pertussis toxin coating solution was then discarded, and the tubes were washed with 0.5 ml of buffer A. The tubes were then washed two times with 0.5-ml portions of 1.0% BSA in buffer A. 0.5 ml of BSA-buffer A solution was added, and the tubes were incubated at room temperature for at least 1 h. This final wash solution was discarded, and 100 pl of 0.1% BSA-buffer A containing unlabeled aglycofetuin derivatives was added to the tubes. The tubes were incubated for an additional 45 min, and '%I-fetuin (approximately ' 2 pmol in 20 pl of 0.1% BSA-buffer A) was added. When present, the concentration of the unlabeled fetuin derivatives was 100 times in molar excess of that of the '"I-fetuin. The binding reaction was allowed to proceed for 60 min at room temperature. The labeled fetuin solutions were then discarded, and the tubes were washed four times with 0.5 ml of 0.1% BSA-buffer A. The amount of bound radioactivity was determined in an LKB Rack Gamma 1270 y counter. The amount of lZ6I-fetuin bound to BSA-saturated polystyrene tubes which had not been precoated with pertussis toxin was 4% of the amount bound to pertussis toxin-coated, BSA-saturated tubes. In addition to unlabeled fetuin, excess unlabeled pertussis toxin was also able to compete for '=I-fetuin binding to pertussis toxin-coated tubes.

RESULTS AND DISCUSSION
graph.
Diagrams of fetuin's N-and 0-linked oligosaccharides are presented in Fig. 1 to illustrate the diversity of structural features which are available for interacting with the pertussis toxin binding site. In addition to the major differences between the two classes of oligosaccharide, there are minor differences which contribute to structural heterogeneity within each class (Edge and Spiro, 1987;Spiro and Bhoyroo, 1974;Takasaki and Kobata, 1986;Townsend et al., 1986). For example, approximately 17% of fetuin's N-linked oligosaccharides differ in the GlcNAc(@l-4)Man(al-3)-linked antenna which contains a Gal glycosidically attached to GkNAc in the (@1,3) instead of (/31,4) linkage. Three 0-linked structures have been identified (Edge and Spiro, 1987;Spiro and Bhoyroo, 1974); a hexa-, tetra-, and trisaccharide. The 0-linked hexasaccharide represents a branched structure in which one of the NeuAc(a2-3)Gal arms is attached to GlcNAc which in turn is (@1,6)-linked to the reducing terminal GalNAc group. The other NeuAc(ry2-3)Gal arm is (/31,3)-linked directly to the terminal GalNAc. Moreover, the GlcNAc (@1,4)-linked to Man and the 0-linked GalNAc sugars may also contain internal (a2,6)-linked NeuAc, the majority of which are (ry2,3)linked to terminal Gal in both oligosaccharide classes.
To investigate which portions of the structures shown in Fig. 1 were important for pertussis toxin binding, exoglycosidase and glycosyltransferase enzymes were used to prepare aglycofetuin derivatives for use in the binding activity assays. The sugar content of the aglycofetuin derivatives was determined by GLC, the results of which are shown in Table I and Fig. 2 accompanied by drawings of the predicted structures of the modified oligosaccharide units. Although neuraminidase and bovine testes @-galactosidase removed 92 and 87% of fetuin's NeuAc and Gal groups, respectively, treatment of asialoagala~fetuin with jack bean /3-N-acetylhexosaminidase caused a decrease of 70% in the hexosamine peak area relative to the peak area of mannose (Table I). However, under the methanolysis conditions used, the N-acetylglucosaminylasparagine bond was probably only partially cleaved (Chambers and Clamp, 1971) and, accordingly, all of the asparagine-linked GlcNAc groups would not have registered on the chromatograms. In addition, only four of the remaining five terminal N-acetylhexosamine groups are in the @-linkage suitable for enzymatic attack. Thus, a less than complete decrease in the area of the N-acetylhexosamine peak was expected and, the data obtained from analysis of the GLC results is consistent with the suggestion that jack bean Nacetylhexosaminidase had released all of the terminal GlcNAc groups from asialoagalactofetuin's N-linked oligosaccharides and the small amount of GlcNAc present in the 0-linked hexasaccharide.
When asialoagalactofetuin was treated with galactosyltransferase, the amount of Gal increased from 6 to 64% of native levels (Table I and Fig. 2). In native fetuin, approximately 75% of the Gal residues are bound in the (/31,4) linkage to GlcNAc on the N-linked oligosaccharides, and approximately 25% are attached (/31,3) to 0-linked GalNAc groups (Fig. 1). In addition, a small number of Gal groups are attached to the 0-linked GlcNAc (Fig. 1). Since galactosyltransferase only adds Gal in the (@1,4) linkage to GlcNAc acceptor groups (Khatra et aL, 1974), the enzyme had apparently attached Gal onto approximately 85% of the available acceptor groups in the asialoagalactofetuin. Therefore, all but 15% of the antennae in the oligosaccharides of the galactosyltransferase-treated asialoag~actofetuin now terminated in Gal(@1-4)GlcNAc, the appropriate acceptor for the 2,6-sialyltransferase. The remaining N-linked antennae terminated in GlcNAc, the (@1-3)-linketi Gal groups found in native fetuin were no longer present, and, except for a small amount of Gal(B1-4)GlcNAc linked to GalNAc (not shown in Figs. 2 and 3), the GalNAc residues were all that remained of the majority of 0linked sugars.
Exposure of the galactosyltransferase-treated asialoagalactofetuin to 2,6-~ialyltransferase in the presence of CMP-NeuAc restored almost 40% of the native level of NeuAc groups (Table I and Fig. 2 1. Structures of fetuin N-and O-linked oligosaccharides as determined by Takasaki and Kobata (1986), Spiro and Bhoyroo (19741, and Edge and Spiro (1987). The relative proportions of each of the structures is also indicated. For clarity, the monosaccharides in each of the oligosaccharide units have been assigned symbols in order to draw the structural diagrams which identify the aglycofetuin derivatives in Figs. 2 and 3. The sialic acids in fetuin were previously identified as NeuAc (Takasaki and Kobata, 1986). The different glycosidic linkages of the important NeuAc and Gal groups are indicated by the orientation of the symbols for these sugars.

23% of all 0-linked
And and, for the present, ignoring the small amount of 0-linked Gal(p1-4)GlcNAc acceptor groups, the data displayed in Table I suggested that the sialyltransfe~ase had added NeuAc to only one of the three potential acceptor sites generated by the galactosyltransferase on the N-linked oligosaccharides. Although a previous study by Weinstein et al. (1982b) indicated that the rat liver 2,6-sialyltransferase used in our study should have been capable of attaching NeuAc to all of the available (/31,4)-linked Gal acceptor groups on the tetraantennary oligosaccharide units of al-acid glycoprotein, van den Eijnden et al. (1980) demonstrated that the 2,6-sialyltransferase from bovine colostrum apparently prefers attaching NeuAc to Manfal-3) antennae in triantennary oligosaccharides such as those found on fetuin. These contrasting results may be due to differences in the acceptor specificities of enzymes from two different sources. However, another explanation could be that the tetraantennary oligosaccharide units of CYI-acid glycoprotein may assume a different spatial configuration than fetuin's triantennary oligosaccharides. Such a difference could also contribute to the preferential action of the rat liver sialyltransferase on one of fetuin's three, N-linked antennae. Alternately, our inability to achieve restoration of NeuAc to all available acceptor sites may have been due to the presence of residual neuramini~se activity in the aglycofetuin preparations.
To resolve this issue we have prepared asialofetuin and asialo-atl-acid g~y c o p~t e i n using a procedure designed to reduce the possibility of neuraminidase contamination (Weinstein et al., 1982a). As an extra precaution, we have performed the resialation reactions in the presence of the neuraminidase inhibitor 2,3-dehydro-2-deoxy-N-acetylneuraminic acid (Reutter et al., 1982). Under these conditions, the rat liver sialyltransferase was able to restore 91 k 12% (n = 4) of the sialic acid to asia1o-al-acid glycoprotein but only 55 z t 6% (n  = 2) to asialofetuin. Therefore, residual neuraminidase activity could have been partially responsible for the low sialation efficiency of galactosyltransferase-treated, asialoagalactofetuin, because we observed a 55% recovery of native fetuin NeuAc levels when neuraminidase activity was inhibited but only a 40% recovery of NeuAc if precautions were not taken to eliminate residual neuraminidase activity. Nonetheless, the observation that 91% of the NeuAc was restored to asialo-alacid glycoprotein whereas only 55% of the NeuAc was restored to asialofetuin is also in agreement with the concept that the rat liver sialyltransferase may transfer NeuAc more efficiently to GalB(1-4)GlcNAc groups on tetraantennary oligosaccharides of asialo-al-acid glycoprotein.

Sugar Groups Involved in Pertussis Toxin Binding to Fetuin
It was suggested previously (Sekura and Zhang, 1985) that the GlcNAc-Man bridge structures in fetuin's N-linked oligosaccharides were important for pertussis toxin binding. Our findings that N-acetylglucosaminidase-treated asialoagalactofetuin retained little activity in both assays supports the earlier suggestion. However, we further concluded that NeuAc was also involved in binding because of the observations that fetuin's binding activity was significantly reduced when the NeuAc groups were removed by neuraminidase treatment (Fig. 3). Moreover, when asialoagalactofetuin was sequentially treated with galactosyltransferase and then 2,6-sialyltransferase (but neither enzyme alone), the resulting fetuin derivative was found to be as active as native fetuin in both our assay systems (Fig. 3). This finding suggested that the (a2,6)linkage of NeuAc is of importance for pertussis toxin binding activity because only one-third of the NeuAc found in native fetuin is linked (a2,6) to Gal and, likewise, our resialated fetuin possessed less NeuAc than native fetuin and yet had full activity in both binding inhibition assays. Moreover, our observation that the activity of asialoagalactofetuin in the polystyrene tube binding assay was actually greater than that of the derivative containing only nonreducing terminal Gal groups, suggested that maximum binding of pertussis toxin requires acetamido-containing sugars in the nonreducing terminal positions of fetuin's oligosaccharide units. Furthermore, in the earlier investigations of pertussis toxin-fetuin interaction (Sekura and Zhang, 1985) it was found that 0-linked oligosaccharides were not involved in pertussis toxin binding activity. Our suggestion of the importance of terminal (a2,6)linked NeuAc groups for binding supports the earlier findings because, regardless of structure, fetuin's 0-linked oligosaccharides contain only (a2,3)-linked NeuAc in the nonreducing terminal position (Fig. 1).
Our results also suggested that pertussis toxin possessed lectin-like properties similar to those of wheat germ agglutinin which binds to terminal GlcNAc and NeuAc residues in complex oligosaccharides (Kronis and Carver, 1985;Wright, 1984). However, wheat germ agglutinin apparently prefers binding to (a2,3)-linked NeuAc groups in oligosaccharides (Kronis and Carver, 1985), whereas our data suggested that pertussis toxin may prefer binding to terminal (a2,6)-linked NeuAc groups. We were not able to test the activity of fetuin derivatives containing the other NeuAc linkage configurations because the necessary sialyltransferase enzymes were not available for the study. However, a panel of sugars and glycosides was screened to further investigate the linkage and sugar specificity of pertussis toxin receptors. The following sugars did not inhibit the goose hemagglutination reaction (unless otherwise noted the final concentration of the sugars was 100 mM): L-glucose, a-D-glucose, j3-D-glucose, glucosamine, N-acetylglucosamine, N-acetylgalactosamine, methyla-D-glucopyranoside, 2-deoxy-D-glucose, glucoheptose, phenyl-j3-D-glucopyranoside (50 mM), arabinose, salicin, galactose, galactosamine, galactitol, mannose, D-mannitol, mannosamine, N-acetylmannosamine, methyl-a-D-mannopyranoside, L-fucose, fructose, L-rhamnose, D-arabinose, D-arabitol, L-xylose, D-xylose, methyl-a-D-xylopyranoside, methyl-j3-Dxylopyranoside, D-ribose, ribitol, 2-deoxy-~-ribose, L-sorbose, sorbitol, lyxose, meso-erythritol, lactose, N-acetyllactosamine, maltose, sucrose, trehalose, melibiose, cellobiose, turanose, raffinose, N-glycolylneuraminic acid, N-acetylneuraminic acid, colominic acid, and tri-N-acetylchitotriose (25 mM). In contrast, chitobiose and neuraminlactose from human milk inhibited pertussis toxin-mediated hemagglutination at a minimum concentration of 50 mM each. The observation that chitobiose inhibited hemagglutination is consistent with the wheat germ agglutin-like activity of pertussis toxin. However, the inhibitory effect was only noticed at concentrations greater than 50 mM. Chitotriose did not inhibit hemagglutination, but it was not possible to test concentrations greater than 25 mM due to the limited solubility of this trisaccharide. Furthermore, the observation that human milkderived neuraminlactose, but not NeuAc alone, lactose, or Nacetyllactosamine inhibited hemagglutination also supported our suggestion of the importance of N-acetyllactosaminelinked NeuAc groups for pertussis toxin binding to fetuin. However, because the neuraminlactose contained both (a2,3)and (a2,6)-linked NeuAc we were unable to use this material to investigate preferential binding of pertussis toxin to one uersus the other NeuAc isomer. Nonetheless, it is important to note that none of the inhibitory sugars prevented the hemagglutination reaction at concentrations approaching the minimum inhibitory concentration of fetuin (0.3 f 0.2 WM; n = 11).
Although the contribution of NeuAc groups to binding appeared to be greater than that of GlcNAc in the hemagglutination inhibition assay (Fig. 3), the "'1 -fetuin binding inhibition studies suggested that the two sugars were of equal structural importance. In the lZ5I-fetuin binding inhibition system, neuraminidase treatment reduced native fetuin's activity by approximately 47%, and, in comparison to the activity of asialoagalactofetuin, N-acetylhexosaminidase caused a 54% decline in activity (Fig. 3). The different behavior of the fetuin derivatives in the two assay systems may indicate differences between the way that pertussis toxin interacts with the oligosaccharide domains of fetuin and receptors present in the membranes of goose erythrocytes.
Pertussis toxin receptor activity of glycoproteins may be modulated by factors other than the structure of the oligosac- Analysis of carbohydrates in exoglycosidase-treated fetuin derivatives by gas-liquid chromatography. Gas liquid chromatography was performed on the trifluoroacetate derivatives of 0-methyl glycosides as described in the text. Only the major isomers of the different monosaccharides are labeled. Peaks (0) were identified as residual buffer components, and the structural diagrams are composites of those shown in Fig. 1. Brackets surrounding the (a2-6)-linked NeuAc symbol in the galactosyltransferase and sialyltransferase-treated derivative indicate that ita assignment to the (al-3)Man-linked antenna is tentatively based on its position in native fetuin and the studies of van den Eijnden et al. (1980). charide sequences. This could provide an explanation for the observation that it required much higher concentrations of neuraminlactose than fetuin to inhtbit pertussis toxin-mediated hemagglutination. The number (valence) of glycosylated sites on glycoprotein receptors, the number of antennae in branched oligosaccharide units, or structural components of the core sugars or peptide sequences may also contribute to binding activity. For example, although the tetraantennary oligosaccharides of al-acid glycoprotein contain the NeuAc (a2-6)Ga1(/31-4)GlcNAc sequence (Montreuil, 1984), we found that the minimum pertussis toxin hemagglutination inhibitory concentration of al-acid glycoprotein was 13.5 f 0.5 PM (n = 4). Conversely, although each fetuin molecule contains only three asparagine-linked oligosaccharides and al-acid glycoprotein contains five to eight (Jeanloz, 1972;Takasaki and Kobata, 1986;Townsend et al., 1986), the average minimum pertussis toxin hemagglutination inhibitory concentration for fetuin was approximately 45 times better (0.3 f 0.2 p~; n = 11) than that of al-acid glycoprotein. Moreover, we have also observed that proteolytic degradation of fetuin also greatly reduced its ability to bind to pertussis toxin (minimum pertussis toxin hemagglutination inhibitory concentration > 10 p~; n = 5). Therefore, we are presently characterizing the pertussis toxin receptors in goose erythrocyte membranes so that it will be possible to determine their relationship to fetuin. We feel that the difference observed in the abilities of al-acid glycoprotein, fetuin glycopeptides, and human neuraminlactose to inhibit hemagglutination provides evidence that the interaction of pertussis toxin with native fetuin (and possibly with receptors on cell surfaces) is dependent on the correct spatial orientation of the important sugar groups rather than on electrostatic interactions with Activity of fetuin derivatives in the goose hemagglutination inhibition (stippled burs) and polystyrene tube binding inhibition (striped burs) assays. The assays were performed as described in the text. The structural diagrams are the same as those shown in Fig. 2. The error bars represent the standard deviation of the mean value of the data obtained from at least three independent trials. negatively charged NeuAc. This is supported by the observation that colominic acid, a polymer of NeuAc, was unable to inhibit the hemagglutination reaction.
The use of exoglycosidase and glycosyltransferase enzymes to probe lectin-like interactions at the molecular level has been extremely useful for examining the receptor specificity of different strains of influenza virus (Carroll et aZ., 1981;Rogers and Paulson, 1983). In this report we demonstrated the broader utility of this approach to obtaining a better understanding of the interaction of pertussis toxin with fetuin. The procedures used for modifying oligosaccharide structures in the present studies should allow us to characterize pertussis toxin receptors while maintaining the preferred molecular environment of the chemical groups which are directly involved in the pertussis toxin binding activity. We have also demonstrated that two glycosyltransferase enzymes can be used sequentially to restore pertussis toxin binding activity to fetuin. This enabled us to prepare a derivative which lacked much of the structural heterogeneity of native fetuin's oligosaccharide units (Figs. 1 and 2) and identify those portions which were important for pertussis toxin binding activity (Fig.  3). Although our data suggested that Gal groups do not appear to play a direct role in pertussis toxin binding, fetuin derivatives containing only (/31,4)-linked Gal may be useful for investigating the receptor specificities of other important lectins.