Closely Related Mammals Have Distinct Asialoglycoprotein Receptor Carbohydrate Specificities*

We recently reported that the rat asialoglycoprotein receptor binds oligosaccharides terminating with sialic acid (Sia) (cid:1) 2,6GalNAc. Despite a high percentage of identical amino acids in their sequences, orthologues of theasialoglycoproteinreceptor(ASGP-R)indifferentmam-malsdifferintheirspecificityforterminalSia (cid:1) 2,6GalNAc. The recombinant subunit 1 of the ASGP-R from the rat (RHL-1 or rat hepatic lectin) and the mouse (MHL-1 or mouse hepatic lectin), which differ at only 12 positions in the amino acid sequence of their carbohydrate recognition domains, binds Sia (cid:1) 2,6GalNAc (cid:2) 1,4GlcNAc (cid:2) 1,2Man-bovine serum albumin and GalNAc (cid:2) 1,4GlcNAc (cid:2) 1,2Man-bovine serum albumin in ratios of 16:1.0 and 1.0:1.0, respectively. Mutagenesis was used to show that amino acids both in the immediate vicinity of the proposed binding site for terminal GalNAc and on the (cid:1) 2 helix that is distant from the binding site contribute to the specificity for terminal Sia (cid:1) 2,6GalNAc. Thus, multiple amino acid sequence alterations in two key locations contribute to the difference in specificity observed for the rat and mouse ASGP-Rs. We hypothesize that the altered specificity of ASPG-R ortho-loguesinsuchevolutionarilycloselyrelatedspeciesreflectsrapidlychangingrequirementsforrecognitionofendoge-nousorexogenousoligosaccharides were ampli- fied using the gene-specific primers MHL-1-F (5 (cid:4) -CGG GAT CCC ATC ATG ACA AAG GAT TAT CAA GAT TTC C-3 (cid:4) ) and MHL-1- (cid:5) R (5 (cid:4) -GCG TGC ACC CCC ATT AGC CTT ATC CAA CTT TGT CTC-3 (cid:4) ), MHL-2-F (5 (cid:4) -CGG GAT CCC ATC ATG GAG AAG GAC TTT CAA GAT ATC C-3 (cid:4) ) and MHL-2- (cid:5) R (5 (cid:4) -GCG TCG ACC CTA GTG GGT GAT GTT CCG TCT CTT TTC G-3 (cid:4) ), respectively. Amplified products were subcloned into pcDNA3.1/V5His-TOPO and sequenced. The resulting cDNAs of subunits 1 and 2 were designated MHL-1V5His and MHL-2V5His, respec- tively. To express the subunits with the V5 epitope and six histidines, the stop codon was changed to encode for a Gly residue followed by a SalI restriction enzyme site. Hepatic Lectin Subunit 1 Amino Acid Changes— Standard molecular techniques were used to make HL-1 amino acid changes. The XcmI site at amino acid residues 156–157 was used to swap CRD-containing regions. Codon-specific mutageneses were performed using Inverse PCR. Changes were confirmed by sequencing. These constructs were expressed in HEK-293T cells, solubilized in Triton X-100, separated on NuPAGE® Bis-Tris gels, and electrophoretically transferred to Immo- bilon P (Millipore). Quantitative Western blots were performed using mouse anti-V5 IgG and rabbit anti-mouse IgG-horseradish peroxidase to determine the expression levels by comparison to known amounts of V5-tagged proteins. Bacterial Expression of CRD Regions— Constructs for

The hepatic asialoglycoprotein receptor (ASGP-R) 1 identified by Van Den Hamer et al. (1) was the first mammalian lectin to be described. A characteristic of the receptor is its ability to rapidly remove glycoproteins from the circulation that have been treated with neuraminidase or mild acid (2). Rapid clearance reflects the specificity of the ASGP-R for terminal ␤-linked galactose (Gal) or N-acetylgalactosamine (GalNAc) residues that are exposed by the removal of terminal sialic acid (Sia). The ASGP-R was subsequently shown to be a hetero-oligomer consisting of two highly homologous subunits, hepatic lectin subunits 1 and 2 (HL-1 and HL-2, respectively) (3)(4)(5). Whereas both subunits are required for the endocytosis of ligands, the carbohydrate binding activity is associated predominantly with HL-1 (6,7). Prior structural and functional studies of the ASGP-R have included the crystallization of human HL-1 (8) and the generation of mice with genetically ablated HL-2 (9) and HL-1 (10). These studies have been informative but have not revealed the identity of endogenous ligands that may be recognized by the ASGP-R in vivo.
We recently reported that, in the rat, the ASGP-R mediates the rapid clearance of bovine serum albumin bearing multiple chemically coupled tetrasaccharides with the sequence Sia␣2,6GalNAc␤1,4GlcNAc␤1,2Man (Sia␣2,6GGnM-bovine serum albumin (BSA)) (11). Because the prolactin-like hormones bearing N-linked oligosaccharides terminating with the sequence Sia␣2,6GalNAc␤1,4GlcNAc are synthesized during the last third of pregnancy in the placenta of the rat (12), they may represent the first examples of endogenous ligands for the ASGP-R. Although rare, the terminal sequence Sia␣2,6GalNAc␤1,4GlcNAc has been found on a number of glycoproteins in addition to the prolactin-like hormones (12) including glycodelin from human amniotic fluid (13), pituitary glycoprotein hormones (14,15), and recombinant human protein C (16). The presence of these structures on multiple glycoproteins raises the possibility that the ASGP-R in other mammalian species may also recognize oligosaccharides terminating with the sequence Sia␣2,6GalNAc␤1,4GlcNAc.
The ASGP-R is expressed by all of the mammals that have been examined (17). The amino acid sequences of HL-1 and HL-2 have high percentages of identical residues across species. For example, mouse HL-1 (MHL-1) and human HL-1 (HHL-1) are 89 and 80% identical to rat HL-1 (RHL-1), respectively, whereas MHL-2 and HHL-2 are 80 and 62% identical to RHL-2, respectively. Since all of the known mammalian HLs share a specificity for terminal Gal or GalNAc (17), the conservation of amino acid sequence among mammalian HL orthologues suggested that, like RHL-1, HLs from other species may also bind terminal Sia␣2,6GalNAc.
The carbohydrate recognition domain (CRD) of HL-1 is the prototypical example of a Ca 2ϩ -dependent CRD (18). A comparison of the binding sites of HL-1 and the CRD domain of the mannose (Man)-binding protein allowed Drickamer and coworkers (19 -21) to propose a model for GalNAc binding to HL-1. Based on this model, residues that determine the specificity for GalNAc versus Man were identified and a form of the Man-binding protein that binds GalNAc was engineered. In this model, the C6 hydroxyl of the bound GalNAc projects away from the protein. This model has not yet been confirmed by the crystallization of HL-1 with a carbohydrate in the binding site; however, it suggests that the binding site of HL-1 could potentially accommodate a sialic acid linked to the C6 hydroxyl of bound Gal or GalNAc.
Here we report that mammalian ASGP-Rs differ in their ability to bind terminal GalNAc and Sia␣2,6GalNAc. Furthermore, multiple changes in the amino acid sequence of HL-1 both in the immediate region of the binding pocket and at more distant sites contribute to the difference in specificity. Taken together, the changes in the amino acid sequence and ligand binding properties of these orthologues raise the possibility that the specificity of the ASGP-R may have changed to accommodate a need to recognize different endogenous and/or exogenous carbohydrate structures.

EXPERIMENTAL PROCEDURES
Materials-Bovine, chicken, mouse, porcine, rat, and rabbit livers were purchased from Pel-Freeze® Biologicals (Rogers, AK). BSA conjugated with an average of 15 trisaccharides with the sequence Sia␣2,6GGnM-BSA and the desialylated GGnM-BSA were prepared as previously described (11). TRIzol reagent and primers were purchased from Invitrogen. Radiolabeling of proteins with [ 125 I], isolation of total liver membrane proteins, and receptor:ligand binding assays were performed as previously described (11). However, the concentration of Ca 2ϩ was increased from 2 to 10 mM in the binding assays.
Preparation of Total Liver Membrane Proteins-Livers were homogenized with a Polytron homogenizer (Brinkmann Instruments) in 5 volumes of buffer containing 25 mM HEPES, 50 mM KCl, 2 mM magnesium acetate, 1 mM dithiothreitol, and 10% (w/v) sucrose. Homogenates were sedimented at 1,500 ϫ g for 5 min, and the resulting supernatants were layered over a 65% (w/v) sucrose cushion prior to sedimentation at 100,000 ϫ g at 4°C for 75 min. The interphase fraction, which contained the total membrane proteins, was collected and stored at Ϫ80°C. Protein concentrations were determined using the Bradford method (Bio-Rad) or the Non-Interfering Protein Assay TM (Geno Technologies, St. Louis, MO).
Hepatic Lectin Subunit 1 Amino Acid Changes-Standard molecular techniques were used to make HL-1 amino acid changes. The XcmI site at amino acid residues 156 -157 was used to swap CRD-containing regions. Codon-specific mutageneses were performed using Inverse PCR. Changes were confirmed by sequencing. These constructs were expressed in HEK-293T cells, solubilized in Triton X-100, separated on NuPAGE® Bis-Tris gels, and electrophoretically transferred to Immobilon P (Millipore). Quantitative Western blots were performed using mouse anti-V5 IgG and rabbit anti-mouse IgG-horseradish peroxidase to determine the expression levels by comparison to known amounts of V5-tagged proteins.
Bacterial Expression of CRD Regions-Constructs for bacterial expression of the CRD region of mouse and rat HL-1 were similar to those described by Iobst and Drickamer (22) in which the amino acid residues from 1 to 149 were deleted. In our constructs, the termination codon was replaced with a glycine-encoding triplet to fuse the CRD with V5 and six histidines. CRD region and the tags were amplified using 5Ј-G-ATGATTAACATATGGCCCGAGCCTTCGTTCGAAGGATCTCCTGCC-CCATCAACTGG-3Ј and 5Ј-CTAGAAGGCACAGTCGAGG-3Ј oligonucleotides. MHL-1V5His and RHL-1V5His served as templates. Amplified product were digested with NdeI and PmeI enzymes and inserted into the NdeI and XhoI (Mung bean nuclease-treated) sites of the pET22b(ϩ) vector. The constructs were expressed in BL21(DE3)pLys (Novagen) and solubilized from inclusion bodies using 7 M guanidine-HCL. Refolding by dilution and dialysis was performed as described previously (22). The constructs were enriched by nickel chelate affinity chromatography on nitrilotriacetic acid-agarose (Invitrogen). The active forms of the constructs were affinity-purified on GalNAc␤-agarose (Calbiochem) as described previously (22). Designations M1-CRDV5His and R1-CRDV5His were assigned for the mouse and rat CRDs, respectively. The binding activities of the CRDs were examined using one pmol of each purified CRD.

Mammals Differ in Their Capacity to Bind Oligosaccharides
Terminating with Sia␣2,6GalNAc and GalNAc-We recently reported that bovine serum albumin bearing multiple chemi-cally conjugated tetrasaccharides terminating with the sequence Sia␣2,6GGnM is rapidly removed from the circulation of the rat following injection (11). Subunit 1of the rat asialoglycoprotein receptor RHL-1 binds Sia␣2,6GGnM-BSA as well as GGnM-BSA and can account for the rapid clearance of both glycoconjugates from the blood. We examined the livers of other mammals to determine whether this specificity is universal or confined to the livers of specific mammals such as the rat. Triton X-100-soluble membrane proteins from rabbit, rat, bovine, mouse, and porcine liver displayed activity for GGnM-BSA (Fig. 1A). A 7-fold range in binding activity per microgram of membrane protein was observed with rabbit having the highest range of activity and pig having the lowest activity. Rabbit and rat liver extracts displayed equal or better activity for Sia␣2,6GGnM-BSA than for GGnM-BSA (Fig. 1B). In contrast, bovine, mouse, and pig liver extracts displayed little binding activity for Sia␣2,6GGnM-BSA.
Since we previously determined that the ASGP-R could account for the Sia␣2,6GalNAc-specific binding activity in rat liver, the difference in the ratio of Sia␣2,6GGnM-BSA:GGnM-BSA binding activity seen in livers from different mammals (Fig. 1C) suggested that orthologues of the ASGP-R in these species may differ in their saccharide specificities. Recombinant subunits 1 (HL-1) and 2 (HL-2) of the ASGP-R from rat, mouse, and human were individually expressed in HEK-293T cells, and their ability to bind GGnM-BSA and Sia␣2,6GGnM-BSA was compared. RHL-1, MHL-1, and HHL-1 each bound GGnM-BSA, whereas HL-2 as observed previously (11) did not show significant binding activity in all three species (data not shown). Using 125 I-labeled Sia␣2,6GGnM-BSA and GGnM-BSA that had identical specific activities, the ratio of Sia␣2,6GGnM-BSA to GGnM-BSA bound by the same amount of HL-1 subunit was 15.7-fold higher for RHL-1 than for MHL-1. Thus, the ratio of Sia␣2,6GGnM-BSA to GGnM-BSA binding activity is higher for both the native and recombinant forms of the rat ASGP-R compared with the native and recombinant forms of the mouse ASGP-R (Fig. 2). Similar to MHL-1, recombinant HHL-1 did not exhibit enhanced binding of Sia␣2,6GGnM-BSA compared with GGnM-BSA (data not shown), indicating that the specificity of the human receptor more closely resembles that of the mouse than that of the rat.
Identification of the Amino Acids Required for Recognition of Sia␣2,6GalNAc␤-When aligned with MHL-1 and HHL-1 ( Fig.  2A), the amino acid sequence of RHL-1 is 88.7% identical to MHL-1 and 78.4% identical to HHL-1 (data not shown), indicating that a limited number of alterations in their amino acid sequences account for the difference in specificity. Chimeric HL-1 subunits were used to establish that changes within the CRD of rat and mouse HL-1 rather than within the cytosolic, transmembrane, and stem regions predominantly account for the difference in specificity. The cytosolic, transmembrane, and stem regions of RHL-1 (residues 1-152) were replaced by the cytosolic, transmembrane, and stem regions of MHL-1 (Fig. 2B, Mut-A) and vice versa (Fig. 2B, Mut-B). Whereas the ratio of Sia␣2,6GGnM-BSA to GGnM-BSA binding was reduced for both constructs, the ratio of 9.2 seen for Mut-A indicated that the differences in the CRD domains of RHL-1 and MHL-1 (amino acids 152-284) account for the greater ratio of Sia␣2,6GGnM-BSA to GGnM-BSA binding by RHL-1.
The amino acid sequences of the CRD domains of RHL-1 and MHL-1 differ at 12 positions. We have divided these positions into three groups as shown in Fig. 2A. Mutagenesis was used to systematically convert amino acids at these positions in MHL-1V5His to those present in RHL-1. Mut-E (Fig. 2), which introduced the C-terminal most three amino acid differences (Group III) of RHL-1 into MHL-1, bound equal amounts of Sia␣2,6GGnM-BSA and GGnM-BSA (a ratio of 1.0) similar to MHL-1. Mutating the Group II amino acids, Gln 233 , Glu 238 , Asn 242 of MHL-1 to Lys 233 , Gly 238 , and Asp 242 (Mut-F), resulted in an HL-1 subunit that bound more Sia␣2,6GGnM-BSA than GGnM-BSA (a ratio of 11.9), effectively converting the specificity of MHL-1 to that of RHL-1.
Other differences in the amino acid sequence of RHL-1 and MHL-1 also contribute to the specificity of RHL-1 for Sia␣2,6GGnM-BSA. The introduction of Glu at position 238 in RHL-1 (Mut-M) did not reduce the ratio of Sia␣2,6GGnM-BSA to GGnM-BSA to that of MHL-1, nor did the introduction of a Gly at position 238 in MHL-1 increase the ratio to that of RHL-1. Thus, a single change within the Group II residues that are in closest proximity to the probable contact residues of HL-1 is not sufficient to account for the difference in specificity seen in RHL-1 and MHL-1.
Rat and Mouse HL-1 CRDs Expressed in Bacteria Retain Their Specificity for Terminal Sia␣2,6GalNAc-Rat and mouse HL-1 CRDs consisting of residues 152-284 and a V5His sequence at the C terminus were expressed in bacteria and isolated by affinity chromatography on GalNAc␤-agarose columns. The affinity-purified products were homogeneous when examined by SDS-PAGE (data not shown) and were active as demonstrated by their ability to bind GGnM-[ 125 I]BSA in the polyethylene glycol precipitation assay (Fig. 3). The bacterially expressed and affinity-purified rat and mouse HL-1 CRDs retain their ability to bind Sia␣2,6GGnM-BSA (Fig. 3); however, the rat HL-1 CRD (R1-CRDV5His) binds 3-5-fold more Sia␣2,6GGnM-[ 125 I]BSA per picomole of CRD than the mouse HL-1 CRD (M1-CRDV5His). The difference in ratio of Sia␣2,6GGnM-[ 125 I]BSA to GGnM-[ 125 I]BSA seen with the bacterially expressed HL-1 CRDs and those expressed in HEK-293T cells may reflect a number of possibilities related to the absence of cytosolic, transmembrane, and stem regions and post-translational modifications such as glycosylation. DISCUSSION We previously reported that the HL-1 subunit of the rat ASGP-R binds terminal Sia␣2,6GalNAc and can account for the rapid clearance of Sia␣2,6GGnM-BSA from the blood following injection (11). The studies we have presented here indicate that orthologues of the ASGP-R display marked differ-ences in their specificity for terminal Sia␣2,6GalNAc. For example, recombinant HL-1 subunits of the rat and mouse ASGP-R differ by 16-fold in their capacity to bind terminal Sia␣2,6GalNAc compared with terminal GalNAc. This is the case, even though 87% of the amino acids in rat and mouse HL-1 is identical. The differential capacity for binding terminal Sia␣2,6GalNAc displayed by extracts prepared from rat, rabbit, mouse, bovine, and porcine liver (Fig. 1) suggests that orthologues of the ASGP-R do not have identical specificities.
The 16-fold difference in the ratio of Sia␣2,6GGnM-BSA: GGnM-BSA binding exhibited by RHL-1 and MHL-1 is of particular interest, because the subunits differ at only 12 positions within the CRD (residues 152-284). An exchange of the cytosolic, transmembrane, and stem regions of rat and mouse HL-1 confirmed that differences within the CRD account for the respective specificities of the rat and mouse ASGP-R for Sia␣2,6GalNAc. Our systematic mutation of the MHL-1 to convert the amino acids at these 12 positions to those of the rat revealed that multiple changes in the sequence of the CRD contribute to the difference in specificity. The introduction of the Group II residues ( Figs. 2A and 4, red-colored residues) produces a receptor that has the specificity of RHL-1 (Fig. 2B,  Mut-F). Modeling studies based on the crystal structure of the HHL-1 CRD suggest that bound GalNAc would be stacked against Trp 243 (colored white in Fig. 4) and that the three and four hydroxyls of the GalNAc would be coordinated by the Ca 2ϩ at site 2 (colored green in Fig. 4) (8). The side chain of either a Gly or Glu at position 238 would be located adjacent to the side chain of Gln 239 that is proposed to coordinate the C4 hydroxyl of the GalNAc (22). However, substitution of Gly 238 in RHL-1 with Glu reduces but does not completely abolish Sia␣2,6GalNAc binding (a ratio of 6.7) and the substitution of Glu 238 in MHL-1 with Gly increases the level of Sia␣2,6GalNAc binding (a ratio of 2.5) but not to the level seen for RHL-1. The presence of a Gly at position 238 is not sufficient to increase Sia␣2,6GalNAc binding by MHL-1 to that seen for RHL-1. Asp 242 , which coordinates the Ca 2ϩ (colored blue in Fig. 4) at site 1 of HL-1, may have an impact on interaction of the sialic acid with the peptide.
Alterations in the sequence that are distant from the proposed GalNAc binding site also have an impact on binding Sia␣2,6GGnM-BSA. For example, the Group I residues ( Fig.  2A) that are located on the surface of the ␣2 helix (blue and yellow residues in Fig. 4) have an impact on binding Sia␣2,6GGnM-BSA. Introduction of three of the five Group I residues at positions 198, 200, and 202 (residues colored blue in Fig. 4) in the presence of the Group II residues reduced the ratio of Sia␣2,6GGnM-BSA to GGnM-BSA binding from 11.9 to 3.6 (Mut-F versus Mut-G). The additional introduction of the remaining two residues at positions 194 and 195 (residues colored yellow in Fig. 4) increased the ratio to 5.9 for Mut-H. Thus, the multiple residues that are present in Group I and form the ␣2 helix must be introduced together to preserve Sia␣2,6GGnM-BSA binding in the presence of the Group II residues. It is notable that, with the exception of Lys 172 (colored aqua in Fig. 4) that enhances the binding of both Sia␣2,6GGnM-BSA and GGnM-BSA, all of the differences in the sequence of RHL-1 and MHL-1 that are visible in the crystal structure are located on the surface of HL-1 that contains the ␣2 helix. In addition, Lys 233 from Group II and His 262 from Group III are also on the surface of HL-1 in close proximity to the Group I residues in the ␣2 helix (Fig. 4). We have not determined how or whether the introduction of Group I residues alone would influence the binding of oligosaccharides. Because the Group I residues are distant from the binding site itself, it is not clear whether they alter interactions with contact residues in the binding site and/or whether they influence the formation of ASGP-R oligomers that are known to exist in vivo. The precise manner in which the HL-1 subunit oligomerizes with itself and HL-2 could have a significant impact on binding terminal Sia␣2,6GalNAc.
It is striking that multiple changes in the sequence of RHL-1 contribute to the increased avidity of RHL-1 for Sia2,6␣GalNAc compared with MHL-1. Thus, it appears that few of the changes in the sequence of RHL-1 and MHL-1 can be viewed as silent alterations with respect to the ligand specificity of the ASGP-R. The different specificities of RHL-1 and MHL-1 may reflect exogenous and/or endogenous selection pressures. For example, the rat produces large amounts of prolactin-like glycoprotein hormones, prolactin-like hormones, that bear N-linked oligosaccharides terminating with Sia2,6␣GalNAc␤1,4GlcNAc␤1, 2Man␣ in the placenta and releases them into the blood (11,12). In contrast, we have not detected either the protein-specific GalNActransferase or glycoproteins bearing terminal Sia2,6␣GalNAc␤1, 4GlcNAc␤1,2Man␣ in the placentas of mice. The difference in the specificity of the rat versus mouse ASGP-R may reflect a requirement to recognize terminal Sia2,6␣GalNAc during pregnancy in the rat. Alternatively, other selection pressures such as infectious or parasitic agents may account for changes in the specificity of the ASGP-R.
Our results indicate that orthologues of the ASGP-R in different mammals cannot be assumed to display identical specificities for carbohydrate moieties terminating with ␤1,4-linked Gal or GalNAc (23) despite the high percentage of identical amino acids in their sequences. Furthermore, glycoproteins bearing terminal ␣2,6-linked sialic acid rather than terminal Gal or GalNAc may prove to be endogenous ligands for the ASGP-R. The multiple amino acid sequence modifications required to account for the altered specificity of the ASGP-R in such evolutionarily closely related mammals may indicate that we are observing a carbohydrate-specific receptor, modulating its specificity to accommodate a need to recognize endogenous or exogenous oligosaccharide structures. Understanding the FIG. 4. Space fill and worm/tube models of the HHL-1 CRD illustrating the locations of amino acids that differ in RHL-1 and MHL-1. The model of the HHL-1 CRD was obtained from Protein Data Bank (code 1DV8) (8). The locations of the amino acids that were mutated from the sequence of MHL-1 to that of RHL-1 are indicated in yellow and blue for Group I residues, red for Group II residues, magenta for Group III residues, and aqua for position 172. In each instance, the first letter indicates the residue in MHL-1, the number indicates the residue location in the sequence, and the second letter indicates the residue in RHL-1. L200V* cannot be seen, because it is located on the inner surface of the ␣2 helix. The GalNAc is proposed to be stacked against Trp 243 shown in white with the six hydroxyls directed through a groove between Gly 238 and Trp 243 . The Ca 2ϩ that is colored green (site 2) coordinates the three and four hydroxyls of GalNAc in the binding site. The site 1 and site 3 Ca 2ϩ ions are shown in blue, and the Cl Ϫ ion is in gray. The locations of the ␣2 helix and the ␣1 helix are indicated in the worm/tube model. The lower structures have been rotated 180 o compared with the upper structures. relationship between the specificity of the ASGP-R and the carbohydrates that are recognized may provide new insight into the role such receptors play in vivo.