Multiple Modes of Binding Enhance the Affinity of DC-SIGN for High Mannose N-Linked Glycans Found on Viral Glycoproteins*

The dendritic cell surface receptor DC-SIGN and the closely related endothelial cell receptor DC-SIGNR specifically recognize high mannose N-linked carbohydrates on viral pathogens. Previous studies have shown that these receptors bind the outer trimannose branch Manα1-3[Manα1-6]Manα present in high mannose structures. Although the trimannoside binds to DC-SIGN or DC-SIGNR more strongly than mannose, additional affinity enhancements are observed in the presence of one or more Manα1-2Manα moieties on the nonreducing termini of oligomannose structures. The molecular basis of this enhancement has been investigated by determining crystal structures of DC-SIGN bound to a synthetic six-mannose fragment of a high mannose N-linked oligosaccharide, Manα1-2Manα1-3[Manα1-2Manα1-6]Manα1-6Man and to the disaccharide Manα1-2Man. The structures reveal mixtures of two binding modes in each case. Each mode features typical C-type lectin binding at the principal Ca2+-binding site by one mannose residue. In addition, other sugar residues form contacts unique to each binding mode. These results suggest that the affinity enhancement displayed toward oligosaccharides decorated with the Manα1-2Manα structure is due in part to multiple binding modes at the primary Ca2+ site, which provide both additional contacts and a statistical (entropic) enhancement of binding.

nose oligosaccharide but lacking all terminal ␣1-2-linked mannoses binds 7-and 4-fold better than mannose to DC-SIGN and DC-SIGNR. The full N-linked high mannose oligosaccharide Man 9 GlcNAc 2 , however, shows much more substantial affinity enhancements (4). These data suggested that the presence of the Man␣1-2Man moieties at the nonreducing termini of high mannose oligosaccharides might provide substantial affinity enhancements, perhaps by interacting with a secondary binding site for this group. The surface glycoproteins of HIV and other enveloped viruses are relatively rich in Man 8 and Man 9 structures (13), so high affinity binding to such glycans contributes to selective interaction of DC-SIGN and DC-SIGNR with these pathogens.
Here, the mechanism by which terminal Man␣1-2Man groups enhance affinity toward DC-SIGN and DC-SIGNR is investigated using synthetic fragments of the full N-linked high mannose structure in binding and structural studies. The data indicate that multiple modes of binding at the DC-SIGN carbohydrate-binding site provide a statistical enhancement of the affinity but do not account for all of the observed affinity differences. The different binding orientations feature contacts between the terminal mannose and different regions of the proteins, which likely provide the remaining component of the increased affinity for larger glycans.

EXPERIMENTAL PROCEDURES
Protein Expression-The DC-SIGN carbohydrate-recognition domain was expressed in Escherichia coli as described (6) and used for cocrystallization with Man␣1-2Man. A similar construct lacking the C-terminal 12 residue extension was used for cocrystallization with Man 6 . Both proteins were purified as described (6).
Synthesis and Purification of Man 6 and Man 9 Oligosaccharides-Compounds Man 9 and Man 6a ( Fig. 1) were prepared analogously to those previously described in the literature (14,15). O-Me protection at the reducing end was chosen to diminish the possible interference of the linker with the binding site of the protein. Compound Man 6b was prepared following the same approach as Man 9 , using methyl 2,3,4-tri-O-benzyl-␣-D-mannopyranoside (16,17) as the core sugar unit. After removal of all protecting groups, the compounds were dialyzed two times each for 12 h against 2 liters of Millipore water and then lyophilized. The Man␣1-2Man disaccharide was purchased from Sigma.
Binding Assays-Solid phase competition binding assays were performed using bacterially expressed CRDs of DC-SIGN and DC-SIGNR, with 125 I-Man-bovine serum albumin employed as the reporter ligand (6). The assays were performed at least twice in duplicate, except that the Man 9 GlcNAc 2 glycan was assayed only once in duplicate because only limited quantities were available. Sugar concentrations were determined using the anthrone reaction (18).
Crystallization and Data Collection-Crystals of DC-SIGN CRD complexed with Man 2 or Man 6b (Fig. 1) were grown at 21°C by hanging drop vapor diffusion (1 l of protein to 1 l of reservoir in a drop for Man 2 and 0.6:0.6 for Man 6b ). The protein solution that gave crystals for the DC-SIGN⅐Man 2 complex contained 10 mg ml Ϫ1 protein, 5 mM CaCl 2 , and 25 mM Man 2 .
The protein solution that gave crystals for the DC-SIGN⅐Man 6b complex contained 5 mg ml Ϫ1 protein, 5 mM CaCl 2 , and 50 mM Man 6b . The reservoir solution for both crystals contained 30% (w/v) polyethylene glycol 3000, 0.2 M NaCl, 0.1 M Tris, pH 7.0. The crystals were transferred to a drop of reservoir solution with added sugar, frozen in liquid nitrogen, and maintained at 100 K during data collection. DC-SIGN⅐Man 6b complex data were measured on an ADSC Q315 CCD detector at beam line 11-1 of the Stanford Synchrotron Radiation Laboratory. DC-SIGN-Man 2 diffraction data were measured on an ADSC Q315 CCD detector at beam line 5.0.2 of the Advanced Light Source. Diffraction data were processed with MOSFLM and SCALA (19) (see Table 1).
Structure Determination-Crystals of both the Man 6b and Man 2 complexes were essentially isomorphous to the previously determined DC-SIGN CRD⅐Man 4 complex (5), even though the latter was obtained using slightly different crystallization conditions. The asymmetric unit contains one copy of the protein⅐ligand complex. Rather than reindexing to allow direct rigid body refinement, the two structures were determined by molecular replacement with the DC-SIGN CRD model from the Man 4 complex (Protein Data Bank ID 1SL4). The Man 2 complex structure was determined with the program MOLREP (19), which gave a correlation coefficient of 70% and the R value of 33% for data to 3 Å. The Man 6b complex was solved with the program COMO (20), which gave a correlation coefficient of 56% and the R of value 31% for data to 3.5 Å. Refinement and map calculations for both structures were performed with CNS (21). The maximum-likelihood amplitude target was used, with bulk solvent and anisotropic temperature factor corrections applied throughout the refinement. As refinement progressed it became clear that the ligand is bound to the site in two alternative, overlapping orientations in both structures. The two conformations were assigned occupancies of 75 and 25% based on the quality of the electron density and refined temperature factors. For each ligand orientation, Fig. 2 shows the F o Ϫ F c electron density calculated from coordinates omitting the indicated orientation but including the other. Water molecules were added to peaks of Ͼ3 in F o Ϫ F c maps that were within hydrogen bond distance to protein, sugar, or other water molecules. The final model of the DC-SIGN CRD⅐Man 2 complex contains residues 253-384 of DC-SIGN, two alternative conformations of the carbohydrate ligand, 3 Ca 2ϩ , and 135 water molecules. The final model of the DC-SIGN CRD⅐Man 6 complex contains residues 253-384 of DC-SIGN, two alternative conformations of the ligand, 3 Ca 2ϩ , and 59 water molecules. The refinement statistics are presented in Table 1.  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6

RESULTS
Relative Affinities of Oligomannose Structures for DC-SIGN and DC-SIGNR-To examine the contribution of the Man␣1-2Man groups present on the termini of high mannose oligosaccharide to DC-SIGN and DC-SIGNR binding, three synthetic oligomannose structures corresponding to fragments of Man 9 GlcNAc 2 were tested in the competition assay. Man 9 is the full 9-mannose structure that would be linked to GlcNAc-GlcNAc-Asn in high mannose N-linked carbohydrates (Fig. 1, green box). Man 6a lacks the three terminal mannoses of Man 9 (Fig. 1, red box). Man 6b is the substructure of Man 9 that lacks the ␣1-3 branch arm (Fig. 1, blue box). These compounds were assayed relative to mannose, and the full Man 9 GlcNAc 2 structure purified from soybean agglutinin was also included for direct comparison. The two Man 6 structures bind similarly, with a 9 -14-fold affinity enhancement relative to mannose, whereas the Man 9 compound binds roughly twice as strongly as the Man 6 glycans. The full Man 9 GlcNAc 2 glycan consistently shows 2-3-fold stronger binding than Man 9 ( Table 2).
Structure of Man 6b Bound to DC-SIGN-Crystallization trials of complexes between Man 9 , Man 6a , and Man 6b with the DC-SIGN CRD yielded cocrystals with the 50 mM Man 6b . The structure of this complex was determined at 2.4 Å resolution. The protein structure is identical to that previously described for complexes with The ligand is bound in two overlapping orientations, in a mixture estimated at 75%, designated the major orientation ( Fig. 3, A and B), and 25%, designated the minor orientation ( Fig. 3, C and D). Of the six mannoses in the compound, only three are visible in the major orientation (Man␣1-2Man␣1-3Man) and two in the minor orientation (Man␣1-2Man). The penultimate ␣1-3-linked mannose that forms one arm of the outer branched trimannose unit (i.e. Man␣1-2Man␣1-3[Man␣1-2Man␣1-6]Man␣1-6Man) binds to the primary Ca 2ϩ site and was similarly observed to bind to the Ca 2ϩ in both the DC-SIGN/Man 4 (5) and DC-SIGN/Man 3 GlcNAc 2 FIGURE 2. Electron density maps for bound ligands. The indicated bound ligand orientation is shown superimposed on the F o Ϫ F c electron density map (green, 2 contour) calculated from a model from which the indicated orientation was omitted but which included the alternative orientation. A, Man 6b major orientation. B, Man 6b minor orientation. C, Man 2 major orientation. D, Man 2 minor orientation.
where ͉F o ͉ ϭ observed structure factor amplitude and ͉F c ͉ ϭ calculated structure factor amplitude for the working and test sets, respectively.
(4) structures. The major orientation corresponds to the arrangement seen in these earlier crystal structures (Fig. 3E). The ␣1-6 branch of the oligosaccharide is not visible, however, which is surprising given the shape complementarity and specific hydrogen bonds between the ␣1-6-linked mannose and Phe 313 , Ser 360 , and other residues in DC-SIGN that were observed in the earlier structures (4,5). The reason for this difference is obscure, especially considering the fact that the crystal is essentially isomorphous to the Man 4 complex (5). In the Man 4 and Man 3 GlcNAc 2 structures, the ␣1-6-linked mannose has higher temperature factors than the ␣1-3-linked mannose, suggesting that it may be more weakly bound. The ␣1-2-linked mannose at the nonreducing terminus directly contacts Val 351 .
In the second, less populated orientation, the same mannose residue is bound to the Ca 2ϩ , but its orientation is reversed by a 180°rotation about a line bisecting the pyranose ring through the C-3-C-4 bond. This rotation exchanges the positions of the 3-and 4-OH groups so that they still form the Ca 2ϩ coordination and hydrogen bonds characteristic of C-type lectin-mannose interactions (Fig. 3, B and D). A similar situation was observed in complexes of mannose-binding proteins with various ligands (22). In this orientation, only two sugars are visible: the mannose at the Ca 2ϩ site and the nonreducing terminal ␣1-2linked mannose, which forms hydrogen bonds with Glu 358 and Ser 360 and which also interacts with the face of the Phe 313 ring (Fig. 3C).
Thus, it appears that the Phe 313 side chain has important roles in the recognition of ligand in either orientation. Because only Man␣1-2Man is visible, it is not possible to distinguish whether these residues correspond to the ␣1-3 or ␣1-6 arms of Man 6b or whether they represent a mixture of both (Fig. 1). An overlay of the major and minor orientations is shown in Fig. 3F.
Structure of Man␣1-2Man Bound to DC-SIGN-To assess whether DC-SIGN might have additional subsites for the Man␣1-2Man residues found at the nonreducing termini of high mannose oligosaccharides, the CRD was cocrystallized with 25 mM Man␣1-2Man. The structure of the disaccharide complex reveals binding only in the principal Ca 2ϩ site; no  other carbohydrate molecules were observed even at low electron density map contour levels. Man␣1-2Man binds at the principal Ca 2ϩ site in two orientations, again related by a 180°rotation about the C-3-C-4 bond bisector. The major orientation is virtually identical to that of the Man␣1-2Man moiety in the minor Man 6b ligand orientation and forms the same contacts with DC-SIGN, including the contact with Phe 313 (Fig. 4, A and B). In the minor orientation, only a single sugar is visible and forms the typical Ca 2ϩ coordination and hydrogen bonds (Fig. 4, C and D). This mannose is oriented identically to the Ca 2ϩ -bound mannose in the major Man 6 orientation. The electron density maps, however, do not make clear whether the sugar bound at the Ca 2ϩ is the reducing or nonreducing end of the disaccharide; it is possible that there is a mixture of the two. In particular, unlike the Man 6b structure, the nonreducing ␣1-2-linked mannose is not visible. The electron density for this sugar is not especially well defined in the Man 6b complex, so the lack of density for this residue in the Man 2 complex could be due to its low occupancy. It is also possible that the favorable interaction with Val 351 seen in the Man 6b structure does not compensate for the loss of entropy required to form this contact in the disaccharide. Models of Man 9 Binding-To assess whether the two modes of binding observed in the Man 6b and Man 2 structures are relevant to a full, 9-mannose oligosaccharide, coordinates for Man 9 GlcNAc 2 obtained by NMR analysis of the free glycan (23) were superimposed on the two orientations of Man 6b . As noted previously (4), superposition of the outer branched trimannose moiety reveals no significant steric clashes between the rest of the oligosaccharide and the protein (Fig. 5, A and B), with only the side chain rotamers of Leu 371 needing adjustment to avoid clashes. Superposition of the second orientation, which places the terminal ␣1-2linked mannose of the outer ␣1-3 arm near Phe 313 , also reveals no clashes with the protein, with the possible exception of interference between Arg 345 and the innermost GlcNAc residue (Fig. 5C). Further modeling, in which the terminal Man␣1-2Man groups of the other arms were superimposed, shows no steric clashes (Fig. 5, D and E). Similar results were obtained with crystallographic coordinates of Man 9 GlcNAc 2 derived from the complex with a neutralizing anti-HIV antibody (24), although in this case some minor adjustments to the carbohydrate were required when the outer branched mannose units were superimposed (data not shown).

DISCUSSION
The structure of the Man 6b ⅐DC-SIGN complex reveals two significantly populated binding modes for the ligand. The major binding mode corresponds to that observed in previous crystal structures, featuring a specific site for the outer branched trimannose unit of high mannose N-linked carbohydrates. In this orientation, additional contacts are formed between a nonreducing ␣1-2-linked terminal mannose and Val 351 . This region of DC-SIGN is also important in binding to fucosylated sugars and a terminal GlcNAc in the GlcNAc 2 Man 3 complex (4). The latter compound binds 17-fold more strongly to DC-SIGN relative to mannose and 4-fold more than the trimannose core, suggesting that the additional interactions contribute significantly to overall specificity. Surprisingly, a second binding orientation was observed in which the mannose at the principal Ca 2ϩ site is reversed, thereby generating new interactions between the nonreducing terminal mannose and the region around Phe 313 . Modeling indicates that this orientation would be able to bind to the protein as part of a full 9-mannose structure (Fig. 5, C-E).
The relevance of the dual binding modes of the Man 6b compound was confirmed by the structure of the Man␣1-2Man disaccharide, which shows the same interactions. In this case, the preferred binding mode leaves the nonreducing end near the Phe 313 site. This probably reflects a different energetic balance of the Man 2 and Man 6 compounds, but in any case it is clear that this binding mode can be significantly populated. The fact that no other binding sites for this ligand were observed suggests that there are no other secondary subsites that interact with terminal Man␣1-2Man disaccharides in larger N-linked high mannose oligosaccharides that might account for enhanced binding to such glycans.
The observation of dual binding modes, each resulting in unique contacts with DC-SIGN, permits a semi-quantitative explanation of the affinity enhancements observed when high mannose structures are decorated with ␣1-2-linked mannose residues at the nonreducing termini. Using [P] and [L] to denote the concentrations of free protein and ligand, respectively, and [PLn] for the concentration of the n th distinct protein⅐ligand complex with an association constant K a(n) , the measured affinity constant is related to the affinity constants of the individual binding modes by the equation , this relationship can be restated as K a(meas) ϭ K a1 (1 ϩ x 2 ϩ . . . ). Because ⌬G (meas) ϭ ϪRTlnK a(meas) ϭ ϪRTlnK a1 Ϫ RTln(1ϩ x 2 ϩ . . . ) ϭ ⌬G 1 Ϫ RTln(1ϩ x 2 ϩ . . . ), ⌬G 1 ϭ ϪRTlnK a1 , and ⌬G n ϭ ϪRTlnK a(n) , it follows that x n ϭ exp (Ϫ(⌬G n Ϫ ⌬G 1 )/RT).
Thus, for two binding modes of equal energy, ⌬G 2 Ϫ ⌬G 1 ϭ 0, x 2 ϭ 1, and ⌬G (meas) ϭ ⌬G 1 Ϫ RTln2, so the ability to bind in two equally energetic modes provides an additional RTln2 of free energy, corresponding to 0.41 kcal mol Ϫ1 at 25°C. For n isoenergetic binding modes the observed association constant will be n times the intrinsic association constant, whereas additional weaker binding modes will increase the association by less than a factor of n. The effect of this statistical factor can be illustrated by comparing the binding of Man␣1-2Man with the biding of mannose. With free 3-and 4-OH groups, mannose could bind in either of two orientations related by a 180°rotation that interchanges the 3-and 4-OH groups, as described above. In the disaccharide, each residue can in principle bind in either orientation, giving a total of four binding modes. Thus, if all of these modes were strictly equivalent, the relative K a for the disaccharide would be twice that of mannose. This argument ignores the possible contribution of an alternative binding arrangement involving the 1-and 2-OH groups, which has only been observed in the case of galactose binding to mannose-binding C-type lectins (25). The predicted ratio is seen for DC-SIGNR, but the ratio is about 4 for DC-SIGN, which probably indicates that there are additional, favorable interactions with DC-SIGN Comparison of an NMR-derived structure of Man 9 GlcNAc 2 (23) with the major and minor Man 6b orientations is shown. The protein is shown in cyan, Man 9 GlcNAc 2 is in green, and Man 6b , Man 4 , or Man␣1-2Man is in yellow. Carbon, nitrogen, oxygen, and calcium are shown in white, blue, red, and green, respectively. A, superposition of Man 9 GlcNAc 2 on the major orientation of Man 6b . For clarity the GlcNAc 2 moiety of Man 9 GlcNAc 2 is not shown. B, superposition of Man 9 GlcNAc 2 on Man 4 (5), which corresponds to the major Man 6b orientation but includes the ␣1-6linked mannose (see text and Fig. 3e). The loop in DC-SIGN-CRD (residues 367-374) that is in the vicinity of the two GlcNAc residues is shown in orange. C-E, superposition of Man 9 GlcNAc 2 onto the major Man␣1-2Man orientation (yellow), corresponding to the minor Man 6b orientation. C, ␣1-3 branch terminus of the outer trimannose core. D, ␣1-6 branch terminus of the outer trimannose core. E, terminus of the ␣1-3 branch of the inner trimannose core. made by the second sugar of the disaccharide. It is also possible that free mannose can bind in only one of two modes, as seen in crystal structures of mannose-binding proteins bound to monosaccharides, which generally show a single orientation rather than a mixture in the binding site (22,25).
For more complex ligands, we can consider the trisaccharide binding mode observed in Man 3 GlcNAc 2 and Man 4 as a relatively high affinity mode. The 5-mannose core of the full Man 9 structure, which lacks all Man␣1-2Man groups, binds to DC-SIGN 7-fold better than mannose and 4-fold better in the case of DC-SIGNR. Man 5 also possesses the inner branched trimannose unit in the core that in the absence of the ␤-linked GlcNAc (4) might also bind. The ability to bind to either the inner or outer branched trimannose units likely explains the enhancement of Man 5 over Man 3 (4). Man 6b binds 14-fold better than mannose to DC-SIGN and 12-fold better than mannose to DC-SIGNR (Table 2). Man 6b lacks the inner trimannose, but the outer branched trimannose binding mode and the "reversed" mode in which the nonreducing terminal Man is bound near Phe 313 are present. The 2-and 4-fold enhancement of Man 6b over Man 5 binding to DC-SIGN and DC-SIGNR, can be accounted for in part by the second binding mode. In the major orientation, the ␣1-2-linked terminal mannose on Man 6b forms additional interactions relative to Man 5 , which might make this orientation inherently stronger. In principle the Man␣1-2Man group present on the termini of both the ␣1-3 and ␣1-6 branches of Man 6b can bind in this second orientation, but they cannot be distinguished in the structure (see "Results"), potentially providing three distinct modes at the principal Ca 2ϩ site (when the outer branched trimannose moiety plus the two termini of the two branches are considered).
Given the unequal occupancies of the two orientations observed in the Man 6b complex, it is likely that an inherently stronger interaction of the major, trimannose binding mode and the statistical effect of the second mode both contribute to the observed affinity enhancements. If the major orientation is of higher affinity than the nonreducing Man␣1-2Man binding mode, the affinity enhancements provided by the latter will be less than a factor of n modes. If we assume that the observed occupancies reflect the relative affinities of the two binding modes, with the estimated 3:1 ratio of occupancy in both the Man 6 and Man 2 structures, x ϭ 1 ⁄ 3, so ⌬⌬G ϭ 0.65 kcal mol Ϫ1 would be the energy difference between these modes. If we further assume that the observed minor mode has an equal mixture of the two different ␣1-2 termini, then the ratios are 3:0.5:0.5, and the energy difference is 1.06 kcal mol Ϫ1 . This result illustrates that small differences in energy caused by differences in contacts combined with entropy losses caused by conformational immobilization can give rise to preferred binding orientations and likely explains why all possible modes are not allowed even though the binding site requires only vicinal, equatorial OH groups for Ca 2ϩ ligation.
The affinities of Man 6a for DC-SIGN and DC-SIGNR are enhanced to a similar extent as for Man 6b . In this case, the outer branched trimannose unit is present, but no Man␣1-2Man moieties are appended to these branches. However, two Man␣1-2Man groups present on the ␣1-3 arm of the inner branched trimannose would provide two more binding modes.
It is also possible that the inner branched trimannose unit could bind. Thus, this compound would appear to have a similar number and kind of binding modes as Man 6b , despite their different covalent structures. In the full Man 9 structure, the inner and outer branched trimannose units are present, as well as the Man␣1-2Man groups attached to the branches. If we assume that the Man␣1-2Man␣1-2Man on the ␣1-3 branch of the inner trimannose structure can provide two more modes, we have a total of six modes, which would explain its further enhancement relative to Man 6 .
Although this analysis makes several assumptions about which modes of binding might or might not occur, it is clear that the ability of high mannose oligosaccharides to interact with DC-SIGN and DC-SIGNR in multiple orientations can give rise to statistical affinity enhancements that are consistent with the measured values. Energetic differences among the different binding modes also play an important role in determining the affinity of each compound. Nonetheless, the 2-3-fold increase in affinity displayed by the full Man 9 GlcNAc 2 structure versus Man 9 is difficult to understand. Perhaps the inner GlcNAc residues restrict the conformation of nearby sugar groups such that there is a smaller entropy penalty for binding, or alternatively, novel contacts are formed between these residues and the surface of the protein.
DC-SIGN and DC-SIGNR serve as receptors for HIV and several other enveloped viruses by binding to the high mannose oligosaccharides present on viral surface glycoproteins. The CRD of DC-SIGN specifically recognizes an internal portion of the carbohydrate, namely the outer branched trimannnose unit unique to these carbohydrates. The presence of Man␣1-2Man enhances the affinity of oligomannose toward these receptors, even though by itself this disaccharide binds only slightly more strongly than mannose. The CRD has an intrinsically high affinity for oligomannose structures, and tetramerization likely provides further avidity enhancements for arrays of such structures (6,26). The ability of DC-SIGN and DC-SIGNR to bind high mannose glycans in multiple orientations may facilitate this multivalent binding of clusters of CRDs to glycans displayed in various arrangements on the surface of the virus, as proposed previously for cell surface recognition by mannose-binding proteins (22). There are some parallels with the mechanism by which a neutralizing antibody to HIV, 2G12, binds specifically to the terminal Man␣1-2Man groups present on high mannose carbohydrates (13,27). The binding site of 2G12 appears to recognize specifically a single orientation of Man␣1-2Man present on the nonreducing termini of Man 9 , but at least two of the three branch termini bind to this antibody, which would contribute some statistical enhancement of affinity. High avidity is provided in this case by the unusual domain-swapped dimeric antibody structure, which is proposed to display appropriately spaced binding sites that match the spacing of these structures on the viral surface (24,28).