Selective Sugar Binding to the Carbohydrate Recognition Domains of the Rat Hepatic and Macrophage Asialoglycoprotein Receptors*

Asialoglycoprotein receptors on the surfaces of both hepatocytes and peritoneal macrophages bind terminal galactose residues of desialylated glycoproteins and me- diate endocytosis and eventual degradation of these ligands. The hepatic receptor binds oligosaccharides with terminal N -acetylgalactosamine residues more tightly than ligands with terminal galactose residues, but the macrophage receptor shows no such differential binding affinity. Carbohydrate recognition domains from the macrophage receptor and the major subunit of the hepatic receptor have been expressed in a bacterial sys- tem and have been shown to retain the distinct binding selectivities of the receptors from which they derive. Binding of a series of N -acyl derivatives of galactosamine suggests that the 2-substituent of these sugars in- teracts with the surface of the hepatic receptor with highest affinity binding observed for the N -propionyl derivative. Chimeric sugar-binding domains have been used to identify three regions of the hepatic receptor that are essential for establishing selectivity for N -acetylgalactosamine over galactose. Based on these results and the orientation of N -acetylgalactosamine when bound to an homologous galactose-binding mu- tant of rat serum mannose-binding protein, a fourth region likely to interact with N -acetylgalactosamine has been identified and probed by site-directed mutagenesis. The results of these studies define a binding pocket for the 2-substituent of N -acetylgalactosamine in the hepatic asialoglycoprotein receptor. Asialoglycoprotein receptors have been in both

I-asialoorosomucoid uptake by macrophages in the presence of monosaccharide inhibitors shows that GalNAc and Gal bind equally well (8). Sugar binding studies using the mouse macrophage receptor show a similar pattern, although the exact results depend on what type of assay is used. In a sugar inhibition assay, Gal and GalNAc bind with approximately equal affinity (9), whereas a competitive enzyme-linked immunoabsorption assay suggests that Gal binds with 3-fold higher selectivity than GalNAc (10). In spite of small variations in the different assay protocols, all of the studies indicate that the hepatic receptor binds GalNAc far more effectively than does the macrophage receptor.
A Gal-binding site closely similar to that of the rat hepatic asialoglycoprotein receptor has been created by modification of the CRD of rat serum mannose-binding protein (MBP-A) (11). For high affinity Gal binding, residues Gln 185 , Asp 187 , and Trp 189 must be inserted into the Man-binding framework. A glycine-rich inserted sequence is necessary for the ability to exclude Man. The crystal structure of this mutant, QPDWG, has been solved alone and in complex with Gal and GalNAc (12). Although these studies provide insight into the mechanism of Gal and GalNAc binding to the asialoglycoprotein receptor CRDs, they do not explain differential binding of these two sugars to the hepatic receptor.
The CRDs of RHL-1 and MGR have been produced in a bacterial expression system and the sugar binding properties of these CRDs have been characterized. Amino acid residues likely to be involved in the selective binding of GalNAc to RHL-1 have been identified by analysis of chimeric and mutagenized versions of the CRDs.

EXPERIMENTAL PROCEDURES
Materials-Monosaccharides, glycosides, acid anhydrides, and grade V BSA were products of Sigma. Gal-BSA was obtained from E-Y Laboratories (San Mateo, CA). Na 125 I was purchased from Amersham Corp.
Polystyrene removawell microtiter plates (Immulon 4) were a product of Dynatech. Boehringer Mannheim was the source for isopropyl-␤-Dthiogalactoside. Triton X-100 was purchased from Bio-Rad Laboratories. Restriction and DNA modification enzymes were obtained from New England BioLabs. Reagents for the polymerase chain reaction were purchased from Perkin-Elmer. DNA sequencing was performed using the Sequenase II kit from U. S. Biochemicals. Oligonucleotides were synthesized on an Applied Biosystems 391 DNA synthesizer. Gal-Sepharose was prepared by the divinyl sulfone procedure (13).
Synthesis of Sugar Derivatives-Derivatives of galactosamine were prepared using acid anhydrides available commercially following the method of Horton et al. (14) except the formyl derivative, which was prepared with mixed acetic-formic anhydride (15).
Expression of RHL-1 and MGR-The gene for the rat MGR was amplified using 5Ј and 3Ј primers based on the previously published cDNA sequence (2): 5Ј-GTGTCTCGGTTTCAGTTCAG-3Ј (bases 121-140) and 5Ј-CTACAATAAGCGGCACTCTG-3Ј (bases 1310 -1291). High molecular weight rat liver DNA (16) was used as the template for 25 cycles of 1 min of denaturation at 94°C, 1 min of annealing at 55°C, and 5 min of extension at 72°C using Taq DNA polymerase (17). A BspEI-PstI fragment of the amplified product was subcloned into vector pUC18 and used to isolate exonic fragments that were assembled with synthetic oligonucleotides to create a partial cDNA encoding the CRD.
Modified BamHI-HindIII fragments were inserted into expression vector pT5T, which contains the bacteriophage T7 promotor (see Fig. 1) (18). The pT5T vector was altered by the addition of the synthetic oligonucleotide shown in Fig. 1 at the BamHI site in the polylinker, which causes termination of the T7 Gene 10 product followed by reinitiation to make the CRD. The oligonucleotides used for the liver receptor and the macrophage receptor were identical. The polypeptide produced in each case is extended at the NH 2 terminus by several residues. The first cysteine residue has been changed to serine because it would normally be paired with a cysteine residue further toward the NH 2 terminus of the intact receptor polypeptide. The preceding alanine in the macrophage receptor is changed to isoleucine as well. Chimeric and mutant constructs were created and sequenced in vector pGEM-3 and transferred to the pT5T vector by moving a BamHI-HindIII fragment. The expression construct was transformed into Escherichia coli strain BL21/DE3, in which the T7 RNA polymerase gene is under control of the lac promoter and operator (19).
Mutagenesis Procedures-Mutagenesis was performed on a BamHI-HindIII fragment corresponding to bases 508 -980 of a cDNA for RHL-1 (20) and a BamHI-HindIII fragment corresponding to bases 694-1256 of the cDNA for MGR inserted into the vector, pGEM-3 grown in E. coli strain HB101. Synthetic oligonucleotides or restriction fragments were introduced at appropriate restriction sites (Fig. 2) using standard recombinant DNA techniques (21).
Production of RHL-1 and MGR-The CRDs of RHL-1 and MGR as well as mutants of these two proteins were prepared as described for the CRD of MBP-A (22) with minor modifications. Overnight cultures of bacteria (10 ml) were diluted to 500 ml in LB medium containing 50 g/ml ampicillin and grown at 37°C for 2.5 h before induction with 0.125 mg/ml isopropyl-␤-D-thiogalactoside. After 3 h of further growth at 37°C, bacteria were harvested, and denatured protein was isolated as inclusion bodies following sonication and centrifugation (22). Following solubilization of the protein with guanidine HCl and centrifugation, Triton X-100 was added to a final concentration of 0.5%. The solution was diluted 5-fold with Ca 2ϩ -containing loading buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8, 25 mM CaCl 2 ) prior to dialysis against 6 volumes of loading buffer with 3 buffer changes. After removal of insoluble material by centrifugation for 60 min at 100,000 ϫ g in a Beckman Ti55.2 rotor, the protein was applied to a 1-ml Gal-Sepharose column. Five 1-ml fractions were collected during rinsing with Ca 2ϩ -containing loading buffer followed by five 0.5-ml fractions of EDTA-containing eluting buffer (1.25 M NaCl, 25 mM Tris-Cl, pH 7.8, 2.5 mM EDTA). Fractions were examined by sodium dodecyl sulfate-polyacrylamide (17.5%) gel electrophoresis (23). Fractions containing pure CRD were pooled and adjusted to a final concentration of 25 mM CaCl 2 and approximately 0.1 mg/ml protein for coating of wells for the binding assays.
Competition Binding Assays-Competition assays, in which wildtype and mutant CRDs were immobilized on polystyrene wells, were performed as described previously (24), using 125 I-Gal 34 -BSA as test ligand in the presence or the absence of competing monosaccharides. Each assay was performed in duplicate. A nonlinear, least squares fitting program (SigmaPlot, Jandel Scientific) was used to obtain K i values, corresponding to the concentration of monosaccharide giving 50% inhibition of neoglycoprotein binding according to the equation: where MAX is the total amount of radioactivity bound in the absence of competing monosaccharide. Mean Ϯ standard deviation values for at least three independent assays were used to calculate K i values.
Direct Binding Assay-Polystyrene plates were coated with CRD and blocked with BSA as for the competition assay (24). 125 I-Gal 34 -BSA was diluted with nonradioactive Gal 34 -BSA to a final specific activity of 1-2 ϫ 10 5 cpm/g. Serial dilutions (2-fold) prepared in BSA-loading buffer (1.25 M NaCl, 25 mM CaCl 2 , 25 mM Tris-Cl, pH 7.8, containing 5% BSA) were incubated in wells for 2 h at 4°C. Wells were emptied, rinsed three times with loading buffer at 4°C, and counted in a Packard PRIAS ␥ counter. The data were fitted to an equation for simple, saturable binding superimposed on a linearly increasing background of nonspecific binding: where BKG is the background binding, SLOPE is the linear increase in nonspecific binding, MAX is the saturation level for specific binding, and K D is the concentration of ligand at which half-maximal specific, saturable binding is achieved. Assays were performed in duplicate. Values reported are the mean Ϯ standard deviation for three independent assays, except in certain cases when only a single assay was performed. Molecular Modeling-The coordinates of the Gal-binding mutant of rat serum MBP-A, QPDWG (12), with bound GalNAc (Protein Data Bank 1AFB) were used as a starting point for modeling the binding site in the CRD or RHL-1 using the Insight II program (BioSym Technologies). Residues from the RHL-1 sequence were inserted at the critical positions from the standard rotomer library. No attempt was made to minimize the structure.

Sugar Binding Selectivity of the CRDs from RHL-1 and MGR-
The gene for MGR was amplified from rat genomic DNA using primers from the 5Ј-and 3Ј-untranslated regions that flank the entire protein-coding sequence. The CRD-coding region was then assembled from genomic restriction fragments, and oligonucleotides were designed to repair the intron boundaries. The resultant partial cDNA was identical to that previously sequenced (2). cDNAs for RHL-1 have been previously described (20). The CRD-coding regions of MGR and RHL-1 were inserted after the bacteriophage T7 promotor in the pT5T expression vector (18), as shown in the schematic diagram in Fig. 1. Expressed proteins were purified following a protocol slightly modified from that previously used for MBP-A. The key difference was the addition of Triton X-100 prior to renaturation by dilution and dialysis to prevent aggregation and increase the yield of protein. The CRDs were purified by affinity chromatography on Gal-Sepharose and analyzed by SDS gel FIG. 1. Schematic diagram of expression plasmids for CRDs of RHL-1 and MGR along with predicted protein products. The pT5T expression vector was modified with synthetic oligonucleotides introduced at the BamHI site in the polylinker. The resulting sequence directs termination of the T7 Gene 10 product followed by reinitiation to make the CRD. The polypeptide produced is extended at the NH 2 terminus by the residues in italics. The terminal methionine residue is removed in the bacteria, and the first cysteine residue of each CRD is changed to serine (underlined). Fig. 3, a single species of polypeptide was obtained from each of the expression systems in a yield of 1-5 mg/l of starting bacterial culture.

electrophoresis. As shown in
Solid phase competition binding assays in which CRDs immobilized on polystyrene wells are probed with 125 I-Gal 34 -BSA were used to characterize ligand binding by the bacterially expressed CRDs. As shown in Fig. 4, RHL-1 alone in the absence of RHL-2/3 mimics the highly GalNAc-selective binding of the intact receptor, because GalNAc competes for binding to RHL-1 60-fold more effectively than does Gal. In contrast, MGR does not show this preferential binding. The absolute affinities of the two proteins for Gal 34 -BSA are essentially the same (68 Ϯ 7 nM for RHL-1 and 69 Ϯ 11 nM for MGR), so the difference between the two CRDs in the competition assay reflects much higher affinity of RHL-1 for GalNAc. These results indicate that the different selectivities of the intact hepatic and macrophage receptors are mirrored in the properties of isolated CRDs from RHL-1 and MGR.
The nature of the interaction between GalNAc and RHL-1 was also probed using a series of galactosamine derivatives. The ability of a series of N-acyl derivatives of galactosamine to compete with 125 I-Gal 34 -BSA for binding to the CRD of RHL-1 are compared in Table I. As the length of the side chain increases from formyl to propionyl, the affinity for RHL-1 increases. The propionyl derivative has the highest affinity for RHL-1 of any monosaccharide tested, with larger chains resulting in decreased affinity. These data suggest that there may be a hydrophobic region of the RHL-1 CRD that accommodates small alkyl chains on the 2 substituent of galactosamine.
RHL-1 and MGR Chimeras-Because RHL-1 but not MGR binds GalNAc preferentially, residues in RHL-1 that differ from those in MGR must provide the selectivity for N-acylated galactosamine. The close sequence similarity between the CRDs of RHL-1 and MGR made it possible to construct a series of stable chimeras between them. Comparison of the sugar binding selectivity of these chimeric CRDs with the character- istics of wild-type RHL-1 and MGR CRDs allows identification of residues that determine the distinct binding specificities of the hepatic and macrophage receptors.
Chimeric cDNAs were constructed using restriction sites that are common to both RHL-1 and MGR cDNAs (Fig. 5). The chimeric proteins expressed from these hybrid cDNAs were then tested in solid phase binding competition assays. The results summarized in Table II reveal that all of the 60-fold selectivity for GalNAc can be accounted for in the COOHterminal 81 residues of RHL-1 (amino acids 204 -284). Although the COOH-terminal 81 residues of RHL-1 are necessary for 60-fold preferential binding of GalNAc, smaller effects are observed when shorter segments of RHL-1 are substituted into MGR. Stepwise substitution of portions of the COOH terminus of RHL-1 suggests that at least three regions contribute to high selectivity GalNAc binding. For example, incorporation of just the COOH-terminal 40 residues of RHL-1 into MGR results in a marginal increase in GalNAc binding selectivity. A more substantial 2-fold increase occurs when a further 25 residues of RHL-1 are included. The major, 20-fold increase in GalNAc selectivity occurs only when residues between 80 and 65 residues from the COOH terminus of RHL-1 are present.
To confirm the importance of each region individually, smaller portions of RHL were used to replace corresponding parts of MGR (Fig. 5). The results of binding studies for these constructs (Table II), combined with the single cross-over data, define three regions that contribute to GalNAc binding. The largest effect maps between residues 204 and 218 of RHL-1. Incorporation of this segment, designated region 1, into MGR increases the selectivity of MGR for GalNAc by 20-fold. Region 2, which comprises residues 219 -243 of RHL-1, produces an additional 2-fold increase in GalNAc-binding affinity. Inclu-sion of region 3, running between residues 244 and 284 of RHL-1, causes a further slight increase in GalNAc binding selectivity.
Importance of Residues in Region 1-Each of regions 1, 2, and 3 was probed individually to determine which amino acids in RHL-1 are responsible for selective GalNAc binding. Within region 1, only three amino acid residues in MGR (Ser 228 , Val 229 , and Val 230 ) differ from the corresponding amino acids in RHL-1 (Pro 206 , Leu 207 , and Asn 208 ). To test the importance of each of these residues, the MGR residues were changed to proline, leucine, and asparagine in a pairwise fashion (Table  III). Although the presence of either proline at position 228 or leucine at position 229 of MGR does not affect GalNAc binding, the presence of asparagine at position 230 is absolutely necessary. Substitution of asparagine at position 230 of MGR was tested alone (mutant MGR/1N) to demonstrate that this residue is sufficient to increase the affinity for GalNAc. As shown in Table III, this single amino acid change results in the full 20-fold increase in GalNAc binding selectivity that has been mapped to region 1.
Because incorporation of asparagine at position 230 in MGR is sufficient for relatively high selectivity GalNAc binding, other amino acids were tested at this position to probe the interaction that leads to such high selectivity. All of the amino acids that were substituted at this position are tolerated better than valine (Table III), which is the amino acid present in MGR. Because alanine can substitute for Asn at position 230, the presence of substituents beyond the ␤-carbon must not be necessary for high affinity GalNAc binding. It is important to note that wild-type MGR, mutant MGR/1N, and the other proteins mutants tested have nearly identical absolute affinities for Gal 34 -BSA (Table III), indicating that the presence of Val 230 in wild-type MGR affects only the extra affinity resulting from interactions with the acetamido group of GalNAc but not other interactions common to Gal and GalNAc.
Model of GalNAc-binding Site-The orientation of GalNAc in the binding site of the QPDWG mutant of MBP-A (12) provides a starting point for interpreting the results of the mutagenesis data described here. Binding to this mutant mimics the behavior of the RHL-1 CRD in several important respects. First, it shows similar selectivity for galactose over mannose (11). Second, NMR data show that upon binding to either the RHL-1 CRD or the QPDWG mutant of MBP-A, resonances corresponding to protons on H5 and H6 of Gal are shifted, suggesting similar stacking of this portion of the sugar against tryptophan in both cases (11). Third, studies with sugar derivatives indicate that substituents on 6-OH do not interact with the CRD of RHL-1 (25), which is reflected in the fact that this portion of the sugar is directed away from the binding site in the crystallographic model of the QPDWG mutant of MBP-A. In this model, the acetamido substituent of GalNAc is oriented toward strand  ␤4 of the CRD as shown in Fig. 6A.
The large effect on GalNAc binding resulting from changes in region 1 is surprising because of its probable position in the CRD. In the model shown in Fig. 6A, the shortest distance between the side chain of Asn 208 of RHL-1 and the N-acetyl group of GalNAc is approximately 7 Å. This observation, together with the fact that many amino acids at this position support high selectivity GalNAc binding, suggests that preferential binding to GalNAc by mutant MGR/1N is unlikely to result from direct contact with the sugar, but the effect of the residue at position 230 could be mediated through interactions with other amino acids in the protein.
Definition of an Additional Region Involved in GalNAc Binding-Based on the crystal structure model, residues were identified that could be within contact distance of both Asn 208 of RHL-1 and the sugar ligand. Because the 20-fold increase in GalNAc selectivity is observed following substitution of a single asparagine residue from RHL-1 in the context of MGR, any residue that mediates interaction with GalNAc is likely to be the same in both RHL-1 and MGR. Based on these two criteria, residues His 256 , Phe 257 , and Thr 258 in RHL-1, corresponding to residues His 278 , Phe 279 , and Thr 280 in MGR, were identified as the most likely candidates (region 4, Fig. 6A). To test the importance of these residues, His 278 and Thr 280 of MGR were replaced with alanine, whereas Phe 279 , which is likely to be buried in the hydrophobic core, was replaced with isoleucine to preserve its hydrophobic character. This position is occupied by isoleucine in the RHL-2/3 polypeptide, which, like MGR, lacks preferential binding to GalNAc. 2 As might be expected from the predicted orientation of Phe 279 toward the interior of the CRD, substitution with isoleucine (mutant MGR/1N/4F3 I) results in only a slight loss of 2 N. I. Ruiz and K. Drickamer, manuscript in preparation.
FIG . 5. Sequences of RHL-1/MGR chimeras and mutants of MGR in regions 1, 2, and 3. The amino acid sequence of the CRD of MGR is shown at the top. For RHL-1 and all chimeric constructs, only the amino acids that differ from MGR are indicated. Approximate locations of restriction sites in the corresponding cDNAs used in making chimeric constructs are indicated by arrows.

TABLE I Binding constants for galactosamine derivatives binding to RHL-1
Inhibition constants for derivatives of galactosamine acylated on the N-2 position were determined using solid phase binding competition assays. The K i for each sugar derivative relative to the K i for GalNAc is reported. selectivity for GalNAc (Table IV). Replacement of Thr 280 with alanine (mutant MGR/1N/4T3 A) has only a 2-fold effect on GalNAc binding. However, substitution of His 278 with alanine results in an almost total loss of GalNAc selectivity. Given the apparent importance of histidine at position 278, it can be suggested that the decrease in GalNAc selectivity in the alanine mutant at positions 280 may result from changes in the position of the histidine residue. To probe the nature of the interaction of His 278 with GalNAc, individual changes were made at this position. None of the amino acids substituted support GalNAc-binding selectivity comparable with that seen when His 278 is present (Table IV).
The importance of histidine in region 4 was further substantiated by incorporating the change His 256 3 Ala in the CRD of RHL-1 rather than in the MGR background. As shown in Table  IV, this change results in 25-fold loss of relative affinity for GalNAc without detectable change in the absolute affinity for Gal 34 -BSA. This finding confirms the importance of His 256 in RHL-1 and verifies that the high affinity GalNAc-binding site created in the context of MGR accurately models the natural site in RHL-1. In all cases, the dramatic effects of changing the histidine residue in region 4 on binding of GalNAc compared with Gal are not accompanied by substantial changes in the absolute affinity for Gal 34 -BSA, demonstrating that although this histidine plays a critical role in establishing the high affinity of RHL-1 for GalNAc, it is does not form part of the binding site for Gal.
Importance of Residues in Region 2-RHL-1 and MGR differ at seven positions within region 2. Based on the relative location of these amino acids predicted from the three-dimensional structure of the QPDWG mutant of MBP-A (Fig. 6A), two of the seven amino acids, Arg 236 and Gly 238 in RHL-1, were judged likely to be located near the sugar-binding site. These residues correspond to Ala 258 and Lys 260 in MGR. To test their importance, Ala 258 and Lys 260 were changed pairwise and individually in mutant MGR/1N. Competition binding assay data (Table V) show that both arginine at position 258 and glycine at position 260 of MGR (mutant MGR/1N/2A3 R, 2K3 G) are necessary to increase selectivity for GalNAc 2-fold. The side chain of arginine has the potential for hydrogen bonding to the acetyl group of the sugar, but substitution of this residue alone confers only a marginal increase in selectivity for GalNAc, whereas substitution of glycine at position 260 has no effect on its own. It is possible that glycine at position 260 plays a role in the correct positioning of arginine at position 258, because the presence of glycine may alter the structure of the loop in which these two residues are located (Fig. 6A).
Residue His 278 was changed to alanine in mutant MGR/1N/ 2A3 R,2K3 G, resulting in the loss of high selectivity GalNAc binding as expected. However, the effect on GalNAc binding seen in the presence of the region 2 changes alone is still evident, because mutant MGR/1N/2A3 R,2K3 G/4H3 A binds GalNAc with at least 2-fold higher selectivity than does wildtype MGR CRD (Table V). These data suggest that the effects of the residues in region 2 are independent of the interactions of His 278 in RHL-1.
Importance of Residues in Region 3-Within region 3, five amino acid residues differ between RHL-1 and MGR. Based on  Fig. 5. Dissociation constants and relative inhibition constants were determined using the solid phase binding assay. K i values for GalNAc relative to the K i for Gal are reported along with dissociation constants for Gal 34 -BSA. the model of RHL-1 in Fig. 6A, only residues Thr 259 and His 262 in RHL-1, which correspond to Ser 281 and Arg 284 of MGR, might be relatively close to the sugar-binding site. As shown in Table V, no significant effect of the relative binding of Gal and GalNAc is observed when region 3 of mutant MGR/1N is re-placed with region 3 of RHL-1, suggesting that the presence of amino acids from region 2 as well as region 1 might be necessary to demonstrate an effect of region 3. Therefore, mutant MGR/1N/2A3 R,2K3 G was modified by changing the residues at positions 281 and 284 in region 3 (Fig. 5). Replacing Arg 284 of MGR with histidine has no significant effect on GalNAc binding, but replacing Ser 281 with threonine results in a mutant that binds GalNAc as well as native RHL-1 does. Because Thr 259 of RHL-1 is predicted from the model in Fig. 6A to be more than 10 Å from the bound sugar, it is unlikely to make direct contact. This residue lies adjacent to region 4, so it might have an indirect effect on the position of the critical histine in this region. Alternatively, the fact that region 3 only has an effect in the presence of the region 2 changes suggests that the loops containing these two regions might interact.

DISCUSSION
The data presented indicate that substitution of four amino acids, asparagine at position 230, arginine at position 258, glycine at position 260, and threonine at position 281 of MGR, is sufficient to give GalNAc binding that is comparable with that of native RHL-1. Using the structure of the QPDWG mutant of MBP-A as a scaffold, the relative positions of the amino acid side chains known to be required for high affinity binding of GalNAc to RHL-1 can be modeled as shown in Fig.  6A. Although the conformations of the side chains are hypothetical, a binding surface can be formed around the atoms  The roles of residues in regions 2 and 3 were analyzed by changing the sequence of mutant MGR/1N to correspond with residues of RHL-1 in these regions. Sequences of mutant constructs are summarized in Fig. 5. Binding constants were determined as in Table II. on the acetamido substituent of the sugar, with the relative importance of specific interactions shown schematically in Fig. 6B. The most important interactions documented in the present studies are with His 256 of RHL-1, which is predicted to be positioned approximately halfway between the terminal methyl group of the 2 substituent of GalNAc and the side chain of Asn 208 . The proximity of Asn 208 to His 256 (under 4 Å in the model) is consistent with the hypothesis that the presence of valine at position 208 decreases the affinity of the CRD for GalNAc indirectly by altering the disposition of the histidine residue. The fact that many other side chains are tolerated at position 208 is also consistent with this suggestion.
The approximate distance from the imidazole ring of His 256 to the terminal methyl group of the N-acetyl moiety in GalNAc is also under 4 Å in the model, suggesting that there may be a direct interaction between His 256 and the 2 substituent of Gal-NAc. Such an interaction would be consistent with the observed importance of this amino acid side chain for the binding of GalNAc but not of Gal. From the enhancement in GalNAc binding affinity, it can be calculated that the interaction with His 256 contributes approximately 8 kJ/mol to the free energy of binding GalNAc. The exact nature of the interaction is not obvious from the data available. Because asparagine and glutamine might be expected to form hydrogen bonds similar to those made by the imidazole ring of histidine, the failure of these amino acids to support tight binding of GalNAc argues against hydrogen bonding. Conversely, packing of the methyl group of the N-acetyl moiety against the imidazole ring might be important, but the inability of tyrosine to sustain tight binding does not support this suggestion. These results must be interpreted with caution, because it is possible that some of the changes may result in structural changes in the adjacent region of the CRD, so that asparagine, glutamine, and tyrosine side chains might be positioned differently from the imidazole ring of histidine and thus be unavailable for making the same types of interactions. The only other obvious possible direct contacts are in region 2, because Arg 236 has the potential to form a hydrogen bond with the carbonyl oxygen on the acetyl portion of the sugar, which is located less than 4 Å away. The interaction with Arg 236 contributes approximately 2 kJ/mol to the total free energy of GalNAc binding to RHL-1.
Although these studies do not allow a detailed description of the nature of the interactions between these amino acids and the 2 substituent of GalNAc, they are certainly consistent with the probable orientation of the sugar ligand predicted from the three-dimensional structure of the QPDWG mutant of MBP-A complexed with GalNAc. The amino acids required for selective high affinity binding of GalNAc more or less surround the acetamido group. It will be of interest to determine how this pocket is able to accommodate the various acyl groups with longer alkyl chains (Table I).
It is noteworthy that the GalNAc-binding site involves resi-dues from several different portions of the RHL-1 CRD, in contrast with the relatively local region of the MBP-A CRD that is involved in Man binding. As noted previously, the minor subunit of hepatic asialoglycoprotein receptor, RHL-2/3, shows equal affinity for Gal and GalNAc and thus resembles MGR in its ligand-binding properties. Examination of the RHL-2/3 sequence reveals that of the critical residues in regions 1-4 of RHL-1, none are present. From an evolutionary perspective, the need for amino acid substitutions in at least four regions of the polypeptide to achieve tight binding to GalNAc suggests that this specificity has been selected in RHL-1 to serve an important biological function, although there has been no such selective pressure on RHL-2/3 or MGR. Evaluation of potential naturally occurring ligands for each of these receptors, possibly bearing distinct patterns of terminal GalNAc and Gal residues, will be of interest.