DC-SIGN Binds to HIV-1 Glycoprotein 120 in a Distinct but Overlapping Fashion Compared with ICAM-2 and ICAM-3*

DC-SIGN is a C-type lectin that binds to endogenous adhesion molecules ICAM-2 and ICAM-3 as well as the viral envelope glycoprotein human immunodeficiency virus, type 1, glycoprotein (gp) 120. We wished to determine whether DC-SIGN binds differently to its endogenous ligands ICAM-2 and ICAM-3 versus HIV-1 gp120. We found that recombinant soluble DC-SIGN bound to gp120-Fc more than 100- and 50-fold better than ICAM-2-Fc and ICAM-3-Fc, respectively. This relative difference was maintained using DC-SIGN expressed on three different CD4-negative cell lines. Although the cell surface affinity for gp120 varied by up to 4-fold on the cell lines examined, the affinity for gp120 was not a correlate of the ability of the cell line to transfer virus. Monosaccharides with equatorial 4-OH groups competed as well as D -mannose for gp120 binding to DC-SIGN, regardless of how the other hydroxyl groups were positioned. Disaccharide competitors and glycan chip analysis showed that DC-SIGN has a preference for oligosaccharides linked in an (cid:1) -anomeric configuration. Alanine-scanning mutagenesis of DC-SIGN revealed that highly conserved residues that coordinate calcium (Asp-366) region of DC-SIGN and does not interfere with the ability of DC-SIGN to recognize glycans on gp120 (10). After five additional washes, the bound DC028 was detected by a goat anti-mouse antibody conjugated with HRP. For increased sensitivity, HRP activity was detected using a chemiluminescence substrate, SuperSignal ELISA Femto (Pierce). The relative light units were measured using a luminometer (LJL Biosys-tems, Sunnyvale, CA). Generation of DC-SIGN Mutants— Site-directed mutagenesis was carried out using the QuickChange kit (Invitrogen) according to the manufacturer’s instructions. Briefly, for each mutant, a pair of specific oligonucleotide primers containing the appropriate base pair change is used in a PCR to generate a de novo vector containing the cognate DC-SIGN mutant. Each mutant was verified by DNA sequencing. The expression level of each DC-SIGN mutant was determined by fluores-cence-activated cell sorter analysis using a conformationally independ- ent (DC028) and a conformationally dependent (507) anti-DC-SIGN mouse monoclonal antibody (10).


ICAM-3.
DC-SIGN is a mannose-specific, calcium-dependent (C-type) lectin expressed on dendritic cells (DCs) 1 and certain subsets of macrophages (1)(2)(3). DC-SIGN binds with high affinity to the envelope glycoprotein gp120 of HIV and has been shown in vitro to transfer HIV from monocyte-derived dendritic cells to permissive CD4 ϩ T-cells (2, 4 -6). It is believed that the high affinity interaction of gp120 with DC-SIGN on DCs facilitates HIV infection of CD4 ϩ -permissive T-cells in the lymph nodes during the natural course of dendritic cell migration from the peripheral mucosa to the secondary lymphoid organs. DC-SIGN also binds to the glycosylated envelopes of other viruses such as feline immunodeficient virus (59), simian immunodeficiency virus (7)(8)(9)(10), hepatitis C (11)(12)(13), Ebola (14 -18), cytomegalovirus (19), and Dengue (20,21). Most important, DC-SIGN has been shown to be the receptor responsible for the productive infection of DCs by Dengue virus (20). Recent studies also show that DC-SIGN binds to non-viral pathogens such as the Leishmania pifanoi (22,23), Schistosoma mansoni (22,24), Helicobacter pylori (22), and Mycobacterium tuberculosis (22,25,26). The binding of M. tuberculosis to DC-SIGN on DCs is believed to trigger interleukin-10 secretion and thus compromise the immunostimulatory function of the targeted DCs (27).
In addition to ligands derived from pathogens, DC-SIGN has been shown to bind to two endogenous human ligands, ICAM-2 and ICAM-3. ICAM-2 is highly expressed on endothelial cells and is thought to bind to DC-SIGN on a subset of DC precursors and mediate DC emigration from the blood (28). ICAM-3 is expressed on naive T-cells, and its interaction with DC-SIGN on DCs is thought to play a crucial role in DC-mediated T-cell activation (29,30).
Peptide sequence and structural analysis indicate that DC-SIGN contains the prototypical 110 -140-residue carbohydrate recognition domain (CRD) (31)(32)(33) common to C-type lectins. All known structures of C-type lectin CRDs show significant structural conservation characterized by the presence of two pairs of highly conserved disulfide bonds (one of which joins the N and C termini of the CRDs), a set of conserved amino acids that form the hydrophobic core of the domain, the highly conserved Asp-Glu-Cys (EDC) amino acid motif involved in both calcium coordination and sugar binding, and a highly conserved aspartic residue that serves to coordinate calcium and facilitate in the correct folding of the domain. The CRDs of different C-type lectins can recognize various saccharide ligands. DC-SIGN contains the highly conserved Glu-Pro-Asn (EPN) motif and thus belongs to the mannose-specific family of lectins (31). Nonetheless, the CRDs of C-type lectins can recognize a remarkably broad range of ligands and have also been shown to bind specifically to proteins (34), lipids (35), and inorganic molecules such as calcium carbonate (36). Moreover, some CRDs are able to bind to both protein and saccharide ligands (37,38).
Biochemical and structural studies using synthetic substrates demonstrate that DC-SIGN preferentially binds to oligomannosides such as those found on N-linked high mannose type glycoproteins (30, 39 -41). Indeed, binding studies with envelope glycoproteins gp95 of feline immunodeficient virus (59) and gp120 of HIV-1 (2,40) along with ICAM-2 (28) and ICAM-3 (30) confirm that DC-SIGN binds high mannose glycans on these glycoproteins. Additional carbohydrate profiling has led to the identification of Lewis blood group Ags (Lewis x, Lewis y, Lewis a, and Lewis b) (22) as additional oligosaccharide structures that are specifically bound by DC-SIGN. Furthermore, recent reports (25,27) also identified the highly mannosylated surface lipoglycan lipoarabinomannan of M. tuberculosis as a ligand that specifically binds DC-SIGN.
Recently, a mucosally transmitted pathogenic SHIV variant (SHIV 162P) was shown to bind to DC-SIGN 3-fold better than its parental derivative, which was non-pathogenic and poorly transmissible. This gain in DC-SIGN binding function was mapped to an addition of an N-linked glycosylation site in the V2 loop (42). In addition, highly purified DC-SIGN ؉ DCs from human gut biopsies can bind and transfer HIV 1-100-fold more efficiently than DC-SIGN Ϫ DCs from the same tissue (43). These data suggest that DC-SIGN may play a role in the pathogenesis of mucosal HIV transmission, and the DC-SIGN/ gp120 interface may be a legitimate target for antimicrobial therapeutics. However, because DC-SIGN also binds to endogenous ICAM-2 and ICAM-3, we wished to determine whether there were any differences that distinguish DC-SIGN binding to gp120 from binding to the two endogenous ligands ICAM-2 and ICAM-3.
In this study, we showed that recombinant gp120 binds with much greater affinity to DC-SIGN than ICAM-2 and ICAM-3 in in vitro and cell surface binding assays. The cell surface affinity of DC-SIGN for gp120 varied by up to 4-fold among the cell lines examined, but the cell surface affinity for gp120 did not correlate with the ability of DC-SIGN to transfer virus suggesting that cell-specific cofactors may be involved. Competition experiments using gp120, ICAM-2, and ICAM-3 revealed a preference for saccharide ligands with an equatorial 4-hydroxyl group. More extensive glycan array profiling also indicated that the presence of an ␣-anomeric glycosidic linkage contributed to the high affinity binding of cognate oligosaccharide ligands. Finally, targeted alanine-scanning mutagenesis of the CRD of DC-SIGN identified critical residues that differentiate DC-SIGN binding to each of the three ligands.

EXPERIMENTAL PROCEDURES
Materials-DNA restriction enzymes were obtained from New England Biolabs (Beverly, MA). High fidelity Pfu Turbo DNA polymerase was obtained from Invitrogen, and oligonucleotides were supplied from MWG Biotech (High Point, NC). All chemicals and monosaccharides were purchased from Sigma. Disaccharides and high binding certified 96-well plates were purchased from Fisher. Isopropyl-␤-D-thiogalactoside was obtained from BioVectra (Oxford, CT). Anti-human Fc monoclonal antibodies conjugated with horseradish peroxidase (HRP) and the 1-step Ultra TMB substrate for ELISA were purchased from Pierce. Purified recombinant ICAM-2-Fc and ICAM-3-Fc proteins were purchased from R&D Systems, Inc. (Minneapolis, MN). Recombinant gp120-Fc was produced and purified as described (40).
Cloning and Purification of sDC-SIGN-The entire extracellular domain of DC-SIGN (sDC-SIGN) beginning with Ala-77 and ending with the natural stop codon of DC-SIGN was cloned into pET15b (Novagen, Madison, WI) using flanking PCR primers with NdeI and BamHI sites. The resulting in-frame fusion gene contains an N-terminal His 6 tag followed by the sDC-SIGN gene.
Freshly transformed BL21/DE3 bacteria was inoculated and allowed to reach mid-log phase at 30°C with shaking before induction with isopropyl-␤-D-thiogalactoside at a final concentration of 1 mM. Expres-sion of sDC-SIGN was carried out in 1 liter of Luria broth supplemented with 100 g/ml ampicillin. The culture was incubated further for at least 3 h before being harvested by centrifuging at 4000 ϫ g for 10 min at 4°C. The resulting pellet was resuspended in 10 ml of 100 mM NaH 2 P 4 ,10 mM Tris-HCl, and 6 M guanidine HCl, pH 8, and lysed by repeated sonication (5 bursts of 1-min duration at 4°C). The lysate was then supplemented with 0.01% ␤-mercaptoethanol and incubated at 4°C for 2 h. The lysate was then centrifuged at 150,000 ϫ g for 30 min at 4°C in a Beckman 55.2Ti rotor, and the supernatant was incubated with 800 l of nickel-nitrilotriacetic acid-agarose resin (Qiagen, Valencia, CA) (pre-equilibrated with denaturing buffer) at 4°C overnight. The resin was loaded onto a 30-cm chromatography column, and all subsequent washes were done with a 30-fold resin-volume excess of wash buffer starting with 30 mM Tris-HCl, pH 8, 0.5 M NaCl, 1 mM CaCl 2 , 6 M urea, and 10 mM imidazole. The column was then washed again with same buffer except 15 mM imidazole was used. Successive washes of 30 mM Tris-HCl, pH 8, 0.5 M NaCl in decreasing concentrations of urea starting with 5 M urea were performed to renature the protein. The protein was eluted with 30 mM Tris-HCl, pH 8, 0.5 M NaCl, 1 mM CaCl 2 , and 1 M imidazole. The eluant was dialyzed twice with 500 ml of 30 mM Tris-HCl, pH 8, 1.0 M NaCl, and 1 mM CaCl 2 to remove excess imidazole.
Saccharide Competition Assays-20 l of a 200 nM sDC-SIGN concentrate were coated onto high binding 96-well plates in the presence of 30 mM Tris-HCl, 30 mM NaHCO 3 , and 1 mM CaCl 2 overnight at 4°C (20 l per well). The plates were then blocked in the same buffer as above but supplemented with 5% bovine serum albumin (50 l per well) at 37°C for 3 h. The plates were washed five times with wash buffer (Tris-buffered saline supplemented with 1 mM CaCl 2 and 0.1% Tween). 2.4, 300, and 150 nM of recombinant gp120-Fc, ICAM-2-Fc, and ICAM-3-Fc, respectively, were incubated in parallel with the indicated amounts of the various saccharides at room temperature for 2 h. The plates were then washed five times with wash buffer and incubated with anti-human Fc monoclonal antibody conjugated with HRP (Pierce) for 1 h at room temperature. The plates were then washed five times, and the amount of bound ligand was assessed with 1-step Ultra TMB substrate (Pierce). The colorimetric reading was performed on a spectrophotometer (Dynex Technologies, Chantilly, VA). For every saccharide, each experiment was performed at least three times, each time in duplicate.
Direct Binding Assays-A 200 nM solution of sDC-SIGN concentrate was coated onto high binding 96-well plates as described above. Increasing amounts of purified recombinant ligand were added and incubated at room temperature for 1 h. The plates were washed five times with wash buffer and incubated with anti-human Fc monoclonal antibody conjugated with HRP as above for 1 h at room temperature. The plates were washed five more times, and the binding was assessed by colorimetry as described above. For each ligand (gp120-Fc, ICAM-2-Fc, or ICAM-3-Fc), each experiment was performed at least three times in duplicate. Dissociation constants (K d ) were calculated using Graph-Pad TM Prism software (San Diego, CA). For the differential lectin binding experiments, 0.2 g/ml of the indicated lectins were coated onto streptavidin plates (Pierce) for 2 h. Plates were washed three times with wash buffer (phosphate-buffered saline supplemented with 0.5% Tween and 2% bovine serum albumin). Ligands were added and bound at room temperature for 1 h, followed by three additional washes. HRP-conjugated goat anti-human Fc antibodies were added for 30 min to detect the amount of bound ligand. The activity of HRP was quantified colorimetrically as described above.
Cell Surface Binding Assays-The B-cell line HS Sultan and the monocytic cell line THP-1 stably expressing DC-SIGN were derived from stable integration of a retroviral vector MIGR1/EGFP vector containing DC-SIGN (6). 293T cells expressing DC-SIGN were produced by transient CaPO 4 transfections with pCDNA3-DCSIGN according to standard protocols (6). The cell surface binding reaction was done as described previously (40). The cell surface K d values were calculated using the GraphPad TM Prism software and were obtained by titrating in the amount of the indicated ligand and normalizing the highest mean fluorescent intensity value obtained to 100%.
Saccharide Chip Array-96-Well format of mono-and oligosaccharides was purchased from Glycominds (Israel). Binding assays with sDC-SIGN were performed according to manufacturer's instructions. Briefly, the 96-well plate was washed five times with Tris-buffered saline, 0.1% Tween (wash buffer) and allowed to bind to 20 l of a 200 nM solution of purified recombinant His-tagged sDC-SIGN at room temperature for 1 h. The plate was washed five times with wash buffer and incubated with a mouse anti-DC-SIGN monoclonal antibody (DC028) for 1 h at room temperature. DC028 recognizes the repeat region of DC-SIGN and does not interfere with the ability of DC-SIGN to recognize glycans on gp120 (10). After five additional washes, the bound DC028 was detected by a goat anti-mouse antibody conjugated with HRP. For increased sensitivity, HRP activity was detected using a chemiluminescence substrate, SuperSignal ELISA Femto (Pierce). The relative light units were measured using a luminometer (LJL Biosystems, Sunnyvale, CA).
Generation of DC-SIGN Mutants-Site-directed mutagenesis was carried out using the QuickChange kit (Invitrogen) according to the manufacturer's instructions. Briefly, for each mutant, a pair of specific oligonucleotide primers containing the appropriate base pair change is used in a PCR to generate a de novo vector containing the cognate DC-SIGN mutant. Each mutant was verified by DNA sequencing. The expression level of each DC-SIGN mutant was determined by fluorescence-activated cell sorter analysis using a conformationally independent (DC028) and a conformationally dependent (507) anti-DC-SIGN mouse monoclonal antibody (10).

Recombinant Soluble DC-SIGN Oligomerizes Similarly to DC-SIGN on Dendritic
Cells-To delineate the differences in the binding affinities of gp120, ICAM-2, and ICAM-3 to DC-SIGN, we first measured the dissociation constant (K d ) of the three ligands in an immobilized solid-phase ELISA format. To this end, we expressed the extracellular domain of DC-SIGN (sDC-SIGN) as an N-terminal His 6 -tagged fusion protein.
To determine whether sDC-SIGN multimerizes into tetramers as has been demonstrated previously (39), we fractionated the recombinant protein on a 5-20% sucrose gradient. As can seen in Fig. 1, purified recombinant sDC-SIGN readily formed trimers and tetramers in agreement with previous biochemical and physical studies (39). Most important, the higher ordered oligomers of purified sDC-SIGN corresponded approximately to the oligomeric state of full-length DC-SIGN proteins isolated from immature monocyte-derived dendritic cells (Fig. 1). To our knowledge, this is the first demonstration that full-length DC-SIGN on primary dendritic cells can also exist in the oligomeric state found for recombinant sDC-SIGN.
gp120 Binds to DC-SIGN with Greater Affinity Than ICAM-2 and ICAM-3-In order to facilitate our binding studies, HIV-1 gp120 was produced as an IgG1-Fc fusion protein (40). By using purified sDC-SIGN, we found that gp120-Fc bound to sDC-SIGN with 100-and 50-fold higher affinity than ICAM-2-Fc and ICAM-3-Fc, respectively ( Fig. 2A). We then asked if the differential affinity of DC-SIGN for the three ligands was maintained when DC-SIGN was expressed on the cell surface. We chose three cell lines (THP-1, HS-Sultan, and 293T HEK cells) with no appreciable binding to gp120, ICAM-2, and ICAM-3 in the absence of DC-SIGN (Ref. 2 and data not shown). THP-1 and HS-Sultan cells were retrovirally transduced, and clones were isolated that stably expressed DC-SIGN at similar levels. HS Sultan is a mature B-cell line, and the functionality of DC-SIGN on B-cells has not been established previously. 293T HEK cells were transiently transfected with DC-SIGN. In all three cell lines, much like the in vitro binding studies, gp120-Fc bound to DC-SIGN with much greater affinity than ICAM-2-Fc and ICAM-3-Fc (Fig. 2B). However, the relative differences in binding affinity between ICAM-2-Fc and ICAM-3-Fc to DC-SIGN were cell type-dependent. For example, although ICAM-2 (K d ϭ 6.66 Ϯ 0.88 nM) bound better than ICAM-3 (K d ϭ 17.0 Ϯ 1.74 nM) on THP-1 DC-SIGN ؉ cells, it bound less avidly than ICAM-3 on HS-Sultan DC-SIGN ؉ cells (K d ϭ 57.9 Ϯ 4.09 versus 13.92 Ϯ 1.38 nM, respectively). Whereas gp120 showed the highest affinity binding to all three cell lines, gp120 bound to HS Sultan DC-SIGN ؉ (K d , 0.017 nM) cells 4-fold better than to 293T DC-SIGN ؉ (K d , 0.078 nM) transfectants and the THP-1 DC-SIGN ؉ (K d , 0.068 nM) cells (Fig. 2B). These data suggest that the ligand binding behavior of DC-SIGN may be modulated in a cell type-specific manner, although we cannot for-mally exclude interactions of gp120, ICAM-2, and ICAM-3 with other cell surface proteins.
Recent studies (44) have noted that the efficiency of the ability of DC-SIGN to transfer virus to permissive T-cells varies depending on the cell type on which DC-SIGN is expressed. Because HS Sultan DC-SIGN ؉ cells bound gp120 with greater affinity than THP-1 DC-SIGN ؉ cells and HS Sultan cells also belong to an antigen-presenting cell lineage (as do dendritic cells and THP-1 cells), we tested whether HS-Sultan DC-SIGN ؉ cells would transfer HIV better than THP-1 DC-SIGN ؉ cells. As seen in Fig. 3, HS Sultan DC-SIGN ؉ cells were extremely inefficient in transferring HIV-1 to permissive T-cells (compare with THP-1 DC-SIGN ϩ cells in Fig. 3). This suggest that viral binding and transfer are dissociable functions of DC-SIGN, and that cell-specific co-factors other than those that contribute to the cell surface affinity for gp120 are likely to contribute to the ability of DC-SIGN to facilitate infection in trans.
ICAM-2 and ICAM-3 Contain Less High Mannose Oligosaccharides than gp120 -Because DC-SIGN is a mannose-specific lectin, one possible reason for the poor affinity of ICAM-2 and ICAM-3 for DC-SIGN compared with gp120 is that ICAM-2 and ICAM-3 have fewer high mannose oligosaccharides compared with gp120. Although gp120 has a larger number of potential N-linked glycosylation sites than ICAM-2 and ICAM-3, the type of glycans (complex/hybrid versus high mannose) on each site can only be determined empirically, Therefore, we used different lectins of various specificities to determine the relative amounts of the major types of glycans on these three ligands. In an equilibrium binding assay, we found that Galanthus nivalis agglutinin bound to gp120 (K d ϭ 0.15 mM) with about 150-fold higher affinity compared with ICAM-2 (K d ϭ 22.7 mM) and ICAM-3 (K d ϭ 20.4 mM), respectively (Fig. 4). Because G. nivalis agglutinin is a lectin that reacts most strongly with multiple terminal ␣(1,3)mannose residues (45), the data showed that gp120 contained more high mannose-type carbohydrates than ICAM-2 and ICAM-3. We also found that Datura stramonium agglutinin bound to gp120 (K d ϭ 0.74 mM) with a higher affinity than either ICAM-2 (K d ϭ 76.3 mM) or ICAM-3 (K d ϭ 58.7 mM) (Fig. 4). On mammalian cells, D. stramonium agglutinin binds to terminal GlcNAc moieties if they are not masked by galactose or sialic acid. Thus, gp120 may contain more terminal GlcNAc residues compared with ICAM-2 or ICAM-3. In toto, the lectin binding profiles suggest that gp120 is glycosylated differently from ICAM-2 and ICAM-3 but that the similar K d values of ICAM-2 and ICAM-3 for G. nivalis agglutinin and D. stramonium agglutinin, respectively, indicate that ICAM-2 and ICAM-3 are glycosylated similarly.
Glycan Specificities of DC-SIGN-mediated Binding-To assess further whether there are any differences in the type of glycans on gp120, ICAM-2, and ICAM-3 that mediate DC-SIGN binding, we used various monosaccharides and disaccharides to compete for ligand binding to immobilized sDC-SIGN. Monosaccharides with equatorial 4-hydroxyl groups (e.g. mannose, mannose derivatives, glucose, and fucose) preferentially com- peted for gp120 binding to DC-SIGN over galactose and galactosides, which have axial 4-hydroxyl groups (Table I; Fig. 5). gp120 and ICAM-3 binding to DC-SIGN were effectively competed by both D-mannose (K i , 5.6 and 1.6 mM, respectively) and D-glucose (K i , 8.08 and 2.35 mM, respectively) but not D-galactose (K i , 63.56 and 68.83 mM, respectively), confirming the relative preference of DC-SIGN for binding to equatorial 4-OH groups ( Table I). As a negative control, L-mannose, the stereoisomer of D-mannose, failed to be an effective competitor for gp120 and ICAM-3 binding (see Table I and Fig. 5A). This preference for a free equatorial 4-hydroxyl group closely mimics another mannose-specific C-type lectin, macrophage mannose receptor (46).
For gp120, DC-SIGN binding was also preferentially competed by oligosaccharides because the disaccharide maltose Glc(␣1,4)Glc (K i , 0.27 mM) was a much better competitor than the monosaccharide D-glucose (Table I; Fig. 5B). Finally, by using disaccharide competitors that differ only in the ␣or ␤-positioning of their glycosidic linkage, such as maltose (Glc(␣1,4)Glc) and its anomer cellobiose (Glc(␤1,4)Glc) (K i , 1.43 mM), we found that DC-SIGN may prefer to bind oligosaccharides linked in an ␣-anomeric configuration (Table I, maltose K i ϭ 0.2 mM; cellobiose K i ϭ 1.4 mM). This preference was later confirmed by saccharide chip analysis (see below and Table II). We understand that dissacharides of glucose are not found in mammalian systems and therefore may not represent a physiological ligand for DC-SIGN. However, because our data suggest that there is no significant difference between D-mannose, L-fucose, and D-glucose binding to DC-SIGN (Table I) and that both competition and direct binding assays suggest a preference for an ␣-anomeric configuration in glucose-based disaccharides, we believe that this preference will hold in more detailed future studies using oligomers of mannose.
Like gp120, D-mannose best competed for ICAM-2 and ICAM-3 binding to sDC-SIGN (Table I). Because it took less D-mannose sugars to compete for the binding of the two endogenous ligands than for gp120, the data suggest that there are less D-mannose sugars on ICAM-2 and ICAM-3 than gp120, consistent with the differential lectin binding data shown in Fig. 4. The monosaccharide competition data of ICAM-3 closely followed the trend of gp120, suggesting that DC-SIGN binding to ICAM-3 was also dependent on the equatorial 4-OH group. However, unlike gp120, maltose (K i , 13.0 mM) competed ϳ8and 6-fold less well than D-mannose (K i , 1.6 mM) and D-glucose (K i , 2.3 mM), respectively (Table I). It is plausible that the binding of sDC-SIGN on the sparsely mannosylated ICAM-3 is unsaturated and more flexible via the widely spaced terminal mannose moieties. Consequently, more maltose is needed to saturate all the binding sites in the DC-SIGN oligomer before it is able to fully compete for the sites bound by ICAM-3. Similarly, the disaccharide maltose also did not compete more effectively than the cognate monosaccharides for DC-SIGN binding to ICAM-2. However, it is peculiar that D-glucose is such a poor competitor of ICAM-2 binding because both competition experiments with gp120 and ICAM-3 show that Dglucose is as good a competitor as D-mannose (Table I). This suggests that DC-SIGN binding to ICAM-2 involves unique elements, which may include protein/protein in addition to protein/carbohydrate interactions.
Saccharide Chip Analysis-To gain further insights into the glycan binding specificities of DC-SIGN, we employed a saccharide microarray analysis of mono-and oligosaccharides dis-TABLE I Saccharide competition data for DC-SIGN binding to gp120, ICAM-2, and ICAM-3 Inhibition constants (K i ) for each monosaccharide and disaccharide were determined by a solid-phase competition assay. At least three independent experiments were done for each condition stated, and each experiment was done in duplicate. 95% confidence intervals (95% C.I.) are indicated in parentheses. ND, not done. The K i of each saccharide tested relative to the K i of D-mannose is shown by the ratio of K i saccharide/K i mannose.
gp 120-Fc ICAM-2-Fc ICAM-3-Fc   4. gp120 contains more mannose sugars than ICAM2 and ICAM3. 0.2 g/ml of the indicated biotinylated lectins were coated onto streptavidin plates and increasing amounts of gp120-Fc, ICAM-2-Fc, or ICAM-3-Fc were added. The amount of ligand bound was detected by adding HRP-conjugated anti-human Fc antibodies. The highest optical density value obtained was normalized to 100%, and binding curves and K d values were generated via Graphpad Prism TM . GNA, G. nivalis agglutinin; DSA, Datura stramonium agglutinin. played in a 96-well ELISA plate format. The carbohydrates were covalently coupled via a flexible linker to the surface, and this technology has been used to profile the saccharide binding specificities of carbohydrate-binding proteins such as lectins and antibodies (47)(48)(49). Here we used sDC-SIGN to bind to the immobilized saccharides in an attempt to characterize the glycan binding specificities of DC-SIGN. These direct binding studies are consistent with the competition data presented in Fig. 5 and Table I. Specifically, sDC-SIGN does not bind to any galactose, galactosides, or any various galactose linked to another sugar (Table II and data not shown). Rather, sDC-SIGN exhibited the greatest binding to glucose and mannose moieties, showing a strict specificity for the 4-OH group in an equatorial position (data not shown and see Ref. 39). In this assay, sDC-SIGN also tends to prefer glycosidic bonds linked in the ␣-anomeric configuration (Table II), consistent with the competition data shown in Table I.
Mutational Analysis of DC-SIGN-The competition and binding data suggest that DC-SIGN may bind differentially to gp120, ICAM-2, and ICAM-3. Therefore, we next sought to determine whether specific residues on DC-SIGN differentially contribute to gp120, ICAM-2, and ICAM-3 binding. To this end, we generated alanine scan mutants that correspond to most of the solvent-exposed amino acids implicated in the maintenance of calcium coordination or carbohydrate contacts in the crystal structure of the CRD of DC-SIGN (Fig. 6A). The 16 different mutations created represent various degrees of conservation among C-type mannose-specific lectins. All alleles were expressed in the HEK 293T cell line, and ligand binding was done similarly to that noted in Fig. 2B. A panel of conformationindependent and conformation-dependent anti-DC-SIGN monoclonal antibodies was used to assess the cell surface expression level and gross structural integrity of each DC-SIGN mutant. mAb 507 is a conformation-dependent antibody directed against the CRD and has the ability to block gp120

TABLE II
Oligosaccharide array analysis of DC-SIGN binding specificity Various mono and oligosaccharides were covalently linked on a 96well ELISA plate as described under "Experimental Procedures." The saccharide array was incubated with 200 nM sDC-SIGN, and the amount of bound sDC-SIGN was detected with a mouse monoclonal antibody (DC028) against the repeat region of DC-SIGN. HRP-conjugated anti-mouse antibodies were used as secondary detection agents along with a chemiluminescent substrate. Data are shown in terms of relative light units, and the luminometer used exhibits a 5-log linearity from 10 2 to 10 7 relative light units. The saccharide binding data are consistent with the competition data presented in Table I. A subset of the binding data indicating a preference of DC-SIGN for an ␣-anomeric glycosidic linkage is shown. Binding data to galactose-based saccharides are shown as negative controls.

Saccharide
Relative light units  Table I  binding to DC-SIGN (50), although its exact epitope is not known. mAb DC028 (10) is a conformation-independent antibody directed against the repeat domain of DC-SIGN and should not be directly affected by mutations in the CRD. As seen in Fig. 6B, the expression of wild-type DC-SIGN and its alleles was compared by normalizing the mean channel fluo-FIG. 6. Alanine-scanning mutagenesis reveals differential DC-SIGN binding determinants to gp120, ICAM-2, and ICAM-3. A, the alignment and the amino acid number shown is that of human DC-SIGN. The 16 amino acids that we have mutated to alanines are indicated by stars. Each of the mutated residues are underscored differentially to highlight the degree of sequence conservations among C-type lectins. The degree of conservation is based on sequence alignments of 10 different C-type mannose-specific lectins: human CD23, DC-SIGN, DC-SIGNR, DCIR, endo180, Langerin, mannose-binding lectin, mincle, macrophage mannose receptor, and rat serum mannose-binding protein. B, cell surface expression of 17 DC-SIGN alleles on 293T was assessed by binding with a DC-SIGN-specific and conformation-independent mAb DC028 that recognizes the repeat domain of DC-SIGN. The conformation-dependent mAb 507 that recognizes the CRD of DC-SIGN was also used to monitor whether there was any gross perturbation of the conformation of DC-SIGN. The mean channel fluorescence of each DC-SIGN allele for each antibody (mAb DC028 and mAb 507) was obtained and normalized to that observed for the wild-type (wt) allele, which was set at 100%. The data are shown as the normalized mean Ϯ S.E. of four independent experiments. C, wild-type and the 16 different DC-SIGN mutants were expressed on 293T and allowed to bind to gp120-Fc, ICAM-2-Fc, or ICAM-3-Fc. R-phycoerythrin-conjugated anti-human Fc secondary antibodies were used to detect the amount of bound ligand. The mean fluorescent intensity of each ligand bound to wild-type DC-SIGN was normalized to 100%. This normalized ligand binding data was in turned normalized to the expression level of each DC-SIGN mutant as determined by DC028 staining in B. Thus, if a ligand bound to a particular mutant at 50% of wild-type levels but the expression of the mutant was only 50% of wild-type levels, the expression-normalized binding of the ligand would be 100% of wild-type levels. The data are plotted as the expression-normalized percentage bound of gp120, ICAM-2, and ICAM-3 for each mutant allele compared with wild-type DC-SIGN. rescence obtained for each allele (and for each antibody) to the wild-type expression, which was set at 100%. mAb DC028 will necessarily measure the absolute amounts of cell surface DC-SIGN, whereas mAb 507, whose epitope is unknown, could potentially be affected by a particular mutation. Therefore, the gp120, ICAM-2, and ICAM-3 binding to DC-SIGN and its alleles were normalized to the expression level as determined by mAb DC028 staining (Fig. 6C). In addition, because all the alleles except for P348A were recognized by mAb 507 at 50% or more of the wild-type level, we believe that none of these mutants exhibit gross perturbation of the structure of DC-SIGN. We believe the decreased expression of P348A is because of the absolute decrease in cell surface DC-SIGN as both DC028 and 507 staining revealed the same percent decrease in expression (Fig. 6B).
The observed phenotypes of the 16 mutants fall into several classes. First, as expected, mutations in residues highly conserved in all C-type lectins, Glu-354, Asp-355, and Asp-366, severely compromised binding to all three ligands. Glu-354 coordinates with calcium and interacts with 4-OH of mannose in the co-crystal structures of an oligosaccharide with both DC-SIGN and DC-SIGNR (41). The amino acid Asp-366 of DC-SIGN coordinates extensively with the calcium ion only. Residues Glu-347, Pro-348, and Asn-349 represent the tripeptide sequence that is a highly conserved signature motif for mannose-specific lectins. Both Glu-347 and Asn-349 residues make extensive contacts with the calcium ion and 3-OH of mannose as observed in the DC-SIGN CRD/oligosaccharide co-crystal structure. Most interesting, mutating Glu-347, Pro-348, and Asn-349 to alanine reduced gp120 binding much more substantially than ICAM-2 and ICAM-3 binding. Notably, the P348A mutant appeared to exhibit enhanced binding to the two adhesion molecules (Fig. 6C), although this could be an artificial effect of normalization as P348A was the only mutant that was expressed at significantly lower levels than all the other alleles (Fig. 6B). In sum, mutagenesis of these conserved residues suggests that the primary binding determinants of all three ligands to DC-SIGN were calcium-and mannose-dependent to a certain degree but also that DC-SIGN binds to gp120 in a similar but distinct fashion from ICAM-2 and ICAM-3.
Mutations of amino acids less conserved in C-type mannose lectins show a wide range of informative phenotypes. Mutating residues Asn-311, Arg-345, Val-351, Ser-360, Gly-361, and Asn-362 to alanines showed only a minimal effect on binding to all three ligands (Fig. 6C). Some of these results were somewhat surprising. For example, Asn-311 was postulated in the co-crystal structure, alongside Phe-313, to form a fitting groove for high mannose recognition that specifically discriminates binding against the inner branch point mannose (41) (Fig. 7D), and Ser-360 maintains extensive contacts with Man3 and Man4 in the pentasaccharide co-crystallized with the CRD of DC-SIGN. On the other hand, the mutants G346A and E353A diminished binding to gp120 by at least 50%, while not having a negative effect on ICAM-2 and ICAM-3 binding. Indeed, the G346A mutation appeared to enhance ICAM-2 and ICAM-3 binding by up to 2-fold. Surprisingly, Gly-346 and Glu-353 did not have direct interactions with the pentasaccharide moiety in the co-crystal structure with the CRD of DC-SIGN. Once again, these data suggest that DC-SIGN binding to the two endogenous ligands (ICAM-2 and ICAM-3) may involve unique elements from that required for gp120 binding, and that our mutagenic analysis using the "natural" ligands of DC-SIGN can provide information about the binding behavior of DC-SIGN that is not immediately obvious even from structural data. Finally, the mutant D367A diminished ICAM-2 and ICAM-3 binding by about 50% while enhancing gp120 binding by up to 2-fold, suggesting yet again that DC-SIGN binding to gp120 is qualitatively different from its binding to ICAM-2 and ICAM-3.

DISCUSSION
The present study provides evidence for the molecular determinants of DC-SIGN binding to a viral ligand, gp120, and its two endogenous ligands, ICAM-2 and ICAM-3. First, using purified components in an in vitro equilibrium binding assay, we showed that gp120 binds to sDC-SIGN with nanomolar affinity, 100-and 50-fold better than ICAM-2 and ICAM-3, respectively. When DC-SIGN is expressed on the cell surface, gp120 still bound with relatively much greater affinity than ICAM-2 and ICAM-3 on all three cell types examined. It is interesting to note that the affinity is much higher when binding is done on cell surface-expressed DC-SIGN. This difference in affinity may be due to cellular factor(s) affecting the binding behavior of DC-SIGN and/or the presence of other cognate binding partners that contribute to the total cell surface affinity of examined ligands. We note, however, that all three cell lines were CD4-negative and lack appreciable gp120, ICAM-2, and ICAM-3 binding activity in the absence of DC-SIGN expression.
We also note that the cell surface affinity of DC-SIGN for gp120 does not correlate with its ability to transfer HIV-1 to permissive T-cells. It has been reported that the capacity of DC-SIGN to transmit HIV to permissive T-cells is greater when it is expressed in THP-1 versus 293 cells (44), suggesting the presence of contributory cellular factors to the ability of DC-SIGN to facilitate infection in trans. Here we show that the difference in transmission between these cell lines was not due to the cell surface affinity for gp120 because the measured K d values for the two cell types were similar. Intriguingly, we also show that even on an antigen-presenting cell line (HS-Sultan) with a 4-fold higher affinity for gp120, DC-SIGN was still not able to facilitate virus infection in trans. For the moment, it appears that the ability of DC-SIGN to transfer virus is restricted to a cellular factor(s) common to dendritic cells and THP-1 cells.
Our competition data provide further insights into the molecular specificities of DC-SIGN glycan binding. Unlike the rat mannose-binding protein (51,52) and the macrophage mannose receptor (46), structural and functional data all support the notion that DC-SIGN prefers to bind mannose-or glucosederived oligosaccharides linked in an ␣-anomeric configuration. Because our data indicate that gp120 contains more high mannose type sugars than ICAM-2 and ICAM-3, and it has been shown previously that most of the high mannose moieties in gp120 are linked in the ␣-anomeric configuration (53)(54)(55), it is therefore not surprising that DC-SIGN binds with so much higher affinity to gp120 than to ICAM-2 or ICAM-3 (Fig. 2). Our competition data also indicate that DC-SIGN has a relatively high affinity for L-fucose. This is consistent with published reports of the ability of DC-SIGN to bind to Lewis blood group antigens (22), which are composed of glycosphingolipids, largely defined by a difference in the position of the linkage following a difference in the number of fucose sugars present (56). Indeed, DC-SIGN specifically binds the bacterium H. pycontacts with a mannose moiety in the crystal structure, appear to be dispensable for gp120 binding but not ICAM-2 or ICAM-3 binding. D, the other end of the oligomannose binding valley enclosed by three amino acids Ser-308, Phe-313, and Asn-311 is shown. The arrows point to the three mannose residues and N-acetylglucosamine (GlcNAc5) that line the binding groove of DC-SIGN. The extensive oligosaccharide binding valley is accented in white. B, the highly conserved binding pocket and the amino acid residues that have most drastic effects in binding to all three ligands (i.e. Glu-347 and Asn-349) are shown. Note the extensive sugar and calcium coordination contacts in this binding pocket. C, the residues in our mutagenesis studies that span the middle of the binding groove are shown. Accessory residues such as Ser-360, shown to have lori (22) and the parasitic worm S. mansoni (24) both of which express the Lewis x antigen.
However, it has become increasingly evident that discerning the glycan binding specificities of DC-SIGN is a bit more complex with respect to what glycans it is predicted to bind versus the actual glycan moieties it does bind on its natural ligands. Although DC-SIGN seems to bind to the vector-borne parasite L. pifanoi, the interaction is not inhibitable by mannan (23), which to date has inhibited DC-SIGN binding to all known ligands. It is possible that DC-SIGN binds via another high affinity glycan found on L. pifanoi that is yet to be characterized or that protein/protein interactions are actually involved in binding. Also, our data showed that maltose is a highly effective inhibitor of gp120 binding to DC-SIGN suggesting that dimannosides would be as good or better ligands for DC-SIGN. Yet synthetic glycoclusters harboring dimannoside substrates failed to bind to cells expressing DC-SIGN (57). Nonetheless, convincing evidence shows that the binding of DC-SIGN to M. tuberculosis is due to the dimannoside, Man(␣1,2)Man, cap of the lipoglycan lipoarabinomannan (ManLam) (25,27 Our alanine-scanning mutagenesis of most of the solventexposed amino acids of DC-SIGN CRD reveals several insights into DC-SIGN function. By using the schematic presented in Feinberg et al. (41) (Fig. 7A), we have mutations at the highly conserved end of the "oligosaccharide binding valley" as exemplified by residues Glu-347, Asn-349, Glu-354, Glu-355, and Asp-366 that significantly diminished binding to all three ligands presumably by destroying calcium coordination and hydrogen bonding to the 4-OH of Man2 (Fig. 7B). It is most likely that this one end of the valley contributes the greatest amount of binding energy to the DC-SIGN/ligand interaction because the consensus motif for all mannose-specific and C-type lectins resides here ( Fig. 6A and Fig. 7B). The binding phenotype of P348A mutant is interesting because the homologous residue in rat serum mannose-binding protein has been shown to undergo cis-trans isomerization in the presence and absence of calcium, respectively (58). In rat serum mannose-binding protein, the peptide bond of the highly conserved glutamic acid preceding proline adopts a cis conformation in the presence of calcium, and this is absolutely required in positioning critical amino acids, such as the highly conserved residue Glu-354 in loop 4, for ligation of calcium ions and subsequent substrate binding. In the absence of calcium, the peptide bond adopts a trans conformation and causes dramatic conformational changes in loops 3 and 4 that consequently move side chain atoms by as much as 12 Å (58). Because the P348A mutant affected gp120 binding to DC-SIGN much more significantly than ICAM-2 or ICAM-3 binding, we speculate that the interactions of gp120 with DC-SIGN occur in a more calcium-and mannose-dependent fashion. The same mutant appears to enhance ICAM-2 and ICAM-3 binding when corrected for expression levels, suggesting that the interaction of DC-SIGN with these adhesion molecules may involve unique elements in addition to protein-carbohydrate contacts.
Moving down toward the other end (flanked by residues Asn-311 and Phe-313, see Fig. 7D) in the valley of the binding groove, we found three alanine mutants S360A, G361A, and N362A that have minimal effect on the binding of DC-SIGN to all three ligands (Fig. 6C). This was surprising because in the co-crystal structure these residues exhibit numerous 3-OH and water-hydrogen bonding with the pentasaccharide (Fig. 7C) (41). At best, ICAM 2 binding to the S360A mutant was decreased by about 30%. Ser-360, which packs closely against Phe-313, is interesting because its interactions with two sugar molecules seem to allow for the proposed discriminatory role of Phe-313. It is possible that other interactions mediated by the dense clusters of cognate glycans in the highly mannosylated gp120 can compensate for the slight loss of binding contributed by an "accessory" residue such as Ser-360. We speculate that the sparsely mannosylated ICAM-3 and ICAM-2 may be more sensitive to these mutations and have less "carbohydrate reserve" to compensate for the loss of such binding energy (especially with the S360A mutant). Finally, the crystal structure of DC-SIGN predicts that Asn-311 and Phe-313 play important roles in forming the other end of the binding pocket and contribute to the ligand specificity of DC-SIGN by favoring the binding to the outer branch mannose as a result of the steric clash imposed by Phe-311 on the ␤-anomeric linkage of the inner trimannose branch point (41). Our data indicate that the N311A mutant still binds to all three ligands, suggesting that Phe-313 may play the dominant critical role in high mannose binding of this end of the pocket. Further mutational analysis at this end of the oligomannose binding valley will determine whether this hypothesis is true for which specific glycoprotein ligand.
It is interesting to note that D367A mutant abolishes binding to ICAM-2 and ICAM-3 but enhances gp120 binding. In the crystal structure, Asp-367 resides in the highly conserved end of the pocket and makes extensive water-mediated bonding with the 6th carbon hydroxyl group of Man2. The mutant suggests that Asp-367 is important for ICAM-2 and ICAM-3 because substituting alanine at this position reduced binding by more than 50% and also suggests that Asp-367 may partially obscure more important interactions with gp120 (as changing this residue to alanine enhances gp120 binding). Conversely, the G346A mutant almost abrogated gp120 binding while enhancing ICAM-2 and ICAM-3 binding, suggesting that ICAM-2 and ICAM-3 binding to DC-SIGN may be qualitatively different from gp120. This same residue in DC-SIGNR, a homologue of DC-SIGN, makes water-mediated contacts with the 6th carbon hydroxyl group and may be important for saccharide binding. This observation, coupled with the fact that Dglucose competes more than 100-fold less efficiently than Dmannose for ICAM-2-Fc binding to DC-SIGN, argues once again that ICAM-2 and ICAM-3 binding to DC-SIGN may involve protein/protein in addition to protein/carbohydrate interactions. It is not uncommon that lectins bind to both saccharide and protein determinants concurrently (37,38). Further experimentation is warranted to determine the validity of this hypothesis.
In this study, we show that DC-SIGN binds to HIV-1 gp120, ICAM-2, and ICAM-3 in a distinct but overlapping fashion. We also provide biochemical and genetic evidence that DC-SIGN binding to ICAM-2 is qualitatively different from DC-SIGN binding to gp120 and ICAM-3. Most important, we have identified residues that differentially contribute to binding all three ligands. Delineation of such differences may be useful for future therapeutic intervention targeting the gp120-DC-SIGN interface.