Structural Basis for Selective Recognition of Oligosaccharides by DC-SIGN and DC-SIGNR

See allHide authors and affiliations

Science  07 Dec 2001:
Vol. 294, Issue 5549, pp. 2163-2166
DOI: 10.1126/science.1066371


Dendritic cell specific intracellular adhesion molecule–3 (ICAM-3) grabbing nonintegrin (DC-SIGN), a C-type lectin present on the surface of dendritic cells, mediates the initial interaction of dendritic cells with T cells by binding to ICAM-3. DC-SIGN and DC-SIGNR, a related receptor found on the endothelium of liver sinusoids, placental capillaries, and lymph nodes, bind to oligosaccharides that are present on the envelope of human immunodeficiency virus (HIV), an interaction that strongly promotes viral infection of T cells. Crystal structures of carbohydrate-recognition domains of DC-SIGN and of DC-SIGNR bound to oligosaccharide, in combination with binding studies, reveal that these receptors selectively recognize endogenous high-mannose oligosaccharides and may represent a new avenue for developing HIV prophylactics.

Initiation of a primary immune response requires the interaction of resting T cells with antigen-presenting dendritic cells (1). Initial interaction of T cells with dendritic cells is mediated by the binding of the T cell surface receptor ICAM-3 to a dendritic cell surface receptor denoted DC-SIGN (2). DC-SIGN may also mediate rolling of dendritic cells on endothelium through interactions with ICAM-2, as part of the trafficking of dendritic cells from the blood to the lymphatic system (3). DC-SIGN is a type II membrane protein in which the extracellular domain consists of a stalk that mediates tetramerization (4) and a COOH terminal carbohydrate recognition domain (CRD) that belongs to the C-type (Ca2+-dependent) lectin superfamily (5). Activation of T cells by dendritic cells can be inhibited by antibodies to DC-SIGN, Ca2+ chelators, mannan, or mannose (2), indicating that Ca2+-dependent carbohydrate-binding activity is responsible for the interaction with ICAM-3. DC-SIGNR, a receptor that is 77% identical in sequence to DC-SIGN, is found on liver sinusoidal endothelium and the endothelium of lymph node sinuses and placental villi (6–8).

The presence of dendritic cells greatly enhances the efficiency of infection of CD4+ T cells by human and simian immunodeficiency viruses (HIV and SIV) (9, 10). Antibodies to DC-SIGN significantly reduce the level of infection of T cells co-cultured with dendritic cells, and DC-SIGN binds strongly to the HIV envelope glycoprotein gp120, suggesting that this enhancement is mediated by DC-SIGN (9). DC-SIGN is abundantly expressed on dendritic cells present in the lamina propria of mucosal tissues such as rectum, uterus, and cervix, leading to the proposal that dendritic cells residing at primary sites of HIV exposure capture the virus and deliver it to T cells in lymphatic tissues (9). DC-SIGNR can also bind immunodeficiency virus, suggesting that it may facilitate transmission of HIV across capillaries in the lymph node and at the maternal-fetal boundary (7, 8). Like the dendritic cell–T cell interaction, the interaction of DC-SIGN and DC-SIGNR with HIV is inhibitable by Ca2+ chelators and mannan (7–9).

In order to determine the precise specificity and mechanism of carbohydrate recognition by these receptors, crystal structures of CRDs from DC-SIGN and DC-SIGNR bound to a pentasaccharide (Fig. 1A) were solved [see also supplemental data (11) and Table 1] (Fig. 2A). The DC-SIGNR crystals contain two crystallographically independent molecules, one of which is bound to Ca2+ and the pentasaccharide. The DC-SIGN crystals contain pairs of CRDs cross-linked by the oligosaccharide, in which one monomer forms the same contacts with the oligosaccharide observed for DC-SIGNR, while the partner monomer interacts with the terminalN-acetylglucosamine (GlcNAc) on the α1-3 branch (GlcNAc1) (Fig. 1A).

Figure 1

Oligosaccharide structures. (A) The pentasaccharide co-crystallized with both DC-SIGN and DC-SIGNR. Residue numbers used in the text are shown in parentheses. (B) N-linked high-mannose structure. The full nine-mannose structure (Man9) is shown, but the presence of α1-2 linked mannose at the branch termini is variable. The inner branched trimannose structure Manα1-3 [Manα1-6]Man is shown in the red box, and the outer trimannose structure is shown in the yellow box. (C) Typical N-linked complex carbohydrate. The portion equivalent to the pentasaccharide used in this study is boxed in red. The inner trimannose structure is common to both complex and high-mannose N-linked oligosaccharides. Gal, galactose; GlcNAc, N-acetylglucosamine; Man, mannose; Sia, sialic acid.

Figure 2

Structure of the CRD of DC-SIGN bound to GlcNAc2Man3. (A) Ribbon diagram of the DC-SIGN CRD (blue), with the bound oligosaccharide shown in a ball-and-stick representation (yellow-green, bonds and carbon atoms; red, oxygen; blue, nitrogen). Oligosaccharide residues are shown with the single letter code G for GlcNAc and M for mannose. Large cyan spheres are the three Ca2+. Disulfide bonds are shown in pink. The DC-SIGNR complex is very similar, except that the fourth disulfide connecting the NH2 and COOH termini is not visible in either copy. (B) Rat serum mannose-binding protein bound to a high-mannose oligosaccharide (17). The color scheme is the same as in (A). A total of 228 Å2 of carbohydrate and 166 Å2 of protein surface area is buried in the MBP-oligosaccharide complex, in contrast to the 445 Å2 of carbohydrate and 328 Å2 of protein surface buried in the DC-SIGNR complex.

Table 1

Data collection and refinement statistics. Unit cell parameters are from postrefinement in SCALEPACK (20). Ramachandran plot regions are defined in PROCHECK (21).

View this table:

The DC-SIGN and DC-SIGNR CRDs adopt the typical long-form C-type lectin fold (12) (Fig. 2). The second α helix (α2) (Fig. 2A) is unusually long and lies closer to the rest of the domain than in most C-type lectin-like structures. The COOH-terminal end of the helix packs against the loop connecting β strands 3 and 4, forming a continuous surface that interacts with the bound oligosaccharide (Fig. 2A). In the following text, residue numbers refer to the higher resolution DC-SIGNR structure unless otherwise noted.

The pentasaccharide is exceptionally well defined in both structures [see fig. 1 of supplemental data (11)]. Despite the different crystallization conditions and the way that the carbohydrate cross-links two monomeric CRDs in the DC-SIGN crystals, the conformation of the pentasaccharide is identical to within experimental error [see supplemental data (11)]. The conformations of the four glycosidic linkages are near their average values found in a survey of crystal structures containing these sugars (13) [see table 1 of supplemental data (11)]. Thus, both proteins interact with a common, low-energy conformation of the oligosaccharide. The ω angle of the α1-6 linkage lies slightly outside of the normal range for this angle, which may be due to the interaction between this disaccharide moiety (Man3-Man4) and Phe325.

The internal α1-3–linked mannose (Man2) (see Fig. 1A for residue numbering) binds to the Ca2+ site found in all C-type lectins. The equatorial 3- and 4-OHs each form coordination bonds with the Ca2+ and hydrogen bonds with amino acids that also serve as Ca2+ ligands (Fig. 3, A and B). Such interactions with vicinal hydroxyl groups are hallmarks of C-type lectin-carbohydrate interactions (5). However, the interactions with the DC-SIGN and DC-SIGNR Ca2+ sites are unusual because they involve an internal, rather than a terminal, sugar. In addition to the interactions of the 3- and 4-OHs, the exocyclic C6 packs against Glu366 and the 6-OH forms a water-mediated contact with Asn379 (Asp367 in DC-SIGN) (Fig. 3, A and B). The two mannose residues that form the α1-6 linked branch drape over the solvent-exposed Phe325 and Ser372 (Fig. 3, C and D). The phenylalanine side chain fits into the crevice between Man3 and Man4, where it makes van der Waals contacts with the central Man3. Ser372, which packs against Phe325, forms hydrogen bonds with the 3-OH of Man4 and a water-mediated hydrogen bond with the 2-OH of Man3. The 3-OH of Man4 also forms a water-mediated hydrogen bond with the backbone amide of Ser374.

Figure 3

Interactions of DC-SIGN and DC-SIGNR with GlcNAc2- Man3. (A) Interaction of the α1-3–linked branch with DC-SIGNR. For clarity, the remaining sugar residues are shown schematically. Ca2+coordination bonds are shown as solid black lines; hydrogen bonds, as dashed lines. Large cyan sphere is Ca2+; red, oxygen; blue, nitrogen. (B) Interactions of the α1-3–linked branch with DC-SIGN. The terminal GlcNAc1 forms a cross-link by forming the typical C-type lectin interactions with the principal Ca2+ site of another CRD. For clarity, only the coordination bonds to the Ca2+ (gray) are shown. (C) Interaction of the α1-6–linked branch with DC-SIGNR, including the central (reducing) mannose (Man3). Different portions of the protein backbone are shown in two shades of blue. The interaction of this branch with DC-SIGN is essentially the same. (D) Electrostatic surface representation of the DC-SIGNR CRD, showing the interaction with the bound oligosaccharide. Asn323 and Phe325, which form a shelf that interacts with the Manα1-6Man branch, are indicated. Regions of positive and negative electrostatic potential (10 kT/e level) are shown in blue and red, respectively. (D) was made with GRASP (19).

The terminal GlcNAc residues also form a number of contacts with DC-SIGN and DC-SIGNR. On the α1-3 branch, the acetamido carbonyl oxygen of GlcNAc1 bonds to Ser363 of DC-SIGNR. This position is a valine in DC-SIGN, which participates in van der Waals interactions with GlcNAc1. The contacts between GlcNAc1 with this position are the only major differences in the interactions of these two proteins with the oligosaccharide. In the DC-SIGN crystals, GlcNAc1 also forms a typical C-type lectin Ca2+ coordination and hydrogen bond network at the principal Ca2+ site on the partner monomer in the dimer, thereby cross-linking the two monomers (Fig. 3B). On the other branch, GlcNAc5 forms a number of hydrogen bond and van der Waals interactions with the protein, fitting into a shelf created by Asn323 and Phe325 (Fig. 3, C and D).

The majority of oligosaccharides on gp120 are high mannose structures (14) (Fig. 1B). Previous studies have shown that both DC-SIGN and DC-SIGNR interact particularly effectively with such structures (4). The binding of different types of N-linked oligosaccharides was compared using quantitative inhibition assays as well as glycoprotein and neoglycolipid blots [see table 2 of supplemental data (11)]. The competition data reveal that the GlcNAc-terminated pentasaccharide used in the crystal structures is a ligand for DC-SIGN and DC-SIGNR and has an affinity comparable to a five-mannose structure (Man5). However, though both proteins bind glycoproteins bearing one or more high mannose oligosaccharides, the blotting data show that neither protein binds to complex oligosaccharides attached to glycoproteins (Fig. 1C). Collectively, these data indicate that DC-SIGN and DC-SIGNR are specific for high-mannose N-linked oligosaccharides.

The selectivity of DC-SIGN and DC-SIGNR for high-mannose structures can be understood from the crystal structures and modeling. The two different Manα1-3[Manα1-6]Man moieties present in a model of Man9 (Fig. 1B) derived from nuclear magnetic resonance (NMR) measurements (15) were superimposed on the equivalent portion of the oligosaccharide complexes. Superimposing the inner trimannose branch point reveals that the core GlcNAc residue linked β1-4 to the first mannose clashes with Phe325 (Fig. 4). Modeling reveals steric clashes with the phenylalanine even when the extreme values of the observed torsion angles (13) are used. In contrast, superposition of the outer trimannose branch point (Fig. 1B, yellow) reveals no steric clashes of the Man9 structure with DC-SIGN or DC-SIGNR. Therefore, although these CRDs recognize the trisaccharide Manα1-3[Manα1-6]Man, in the context of N-linked glycans they can only do so when the central mannose is linked in the α anomeric configuration, a structure found only in high-mannose oligosaccharides (compare Fig. 1, B and C). Phe325 plays two essential roles: it forms part of the surface complementary to the shape of the Manα1-6Man moiety (Fig. 3, C and D) and it also provides discrimination between the inner and outer trimannose structures by preventing binding to the inner branch point mannose. Selective binding of DC-SIGN and DC-SIGNR to the outer branched trimannose moiety can explain almost all of their binding characteristics (16).

Figure 4

A phenylalanine prevents binding of the inner trimannosyl core of N-linked oligosaccharides. The DC-SIGNR CRD is shown in cyan, with the Phe325 side chain in a ball-and-stick representation. In yellow-green, the outer branched trimannose structure of Man9 was superimposed on the trimannose structure seen in the DC-SIGNR crystals. For clarity, only the α1-6 branch is shown. In magenta, the mannose residue of a model of the internal Manβ1-4GlcNAcβ1-4GlcNAc moiety of N-linked oligosaccharides was superimposed on the central (reducing) mannose of the trimannose structure. The model was made by setting the torsion angles of the glycosidic linkages to their average values found in an oligosaccharide structure database (13), with adjustments in the torsion angles to overlay the branched trimannose structure precisely. The clash of the first GlcNAc with Phe325 is evident. In both the positions of the α1-3 and α1-6 linked mannose residues linked to the central mannose of the trimannose structure are indicated in the black text.

Mannose-binding proteins (MBPs) are C-type lectins that function in innate immunity by recognizing carbohydrate structures characteristic of pathogens, and they do not bind strongly to mannose-type ligands present on host cell surfaces (5). Although the CRDs of DC-SIGN and DC-SIGNR share 24% sequence identity with rat serum MBP, the mode of binding of oligosaccharides to MBP is quite distinct from the interaction of DC-SIGN and DC-SIGNR with the branched trimannose structure. Each CRD in MBP interacts with a single terminal mannose or GlcNAc residue in an oligosaccharide ligand, and the rest of the oligosaccharide points away from the surface of the protein (17) (Fig. 2B). The architecture of the MBP trimer places the binding sites in the three CRDs far enough apart that MBP binds with high avidity only to the repetitive and dense arrays presented on pathogenic cell surfaces, but not to the more closely spaced terminal mannose residues present on endogenous oligosaccharides (18). In contrast, the interactions of DC-SIGN and DC-SIGNR with endogenous glycans on T cells and the HIV envelope result from high affinity binding to a characteristic internal feature of high-mannose oligosaccharides.

DC-SIGN– and DC-SIGNR–gp120 interactions represent a potential target for anti-HIV therapy aimed at disrupting the DC-virus interaction at primary sites of infection, in order to lower the efficiency of T cell infection. Although the high avidity generated by clustering low-affinity lectin monomers into oligomeric structures has made it difficult to design drugs aimed at disrupting protein-carbohydrate interactions, the unusually high affinity between the monomeric DC-SIGN or DC-SIGNR CRDs and high-mannose oligosaccharides (4) suggests that they may be useful targets. The mechanistic basis of the DC-SIGN– and DC-SIGNR–oligosaccharide interactions presented here provides a starting point for design of such therapeutics, which would attack the viral infection at a novel stage and which could potentially be prophylactic. Lastly, these studies suggest that the interaction of DC-SIGN and DC-SIGNR with endogenous ligands may not be restricted to ICAMs that have been studied to date, but may include other cell surface or soluble glycoproteins with appropriately displayed high mannose oligosaccharides.

  • * To whom correspondence should be addressed. E-mail: bill.weis{at}


View Abstract

Navigate This Article