Structural Requirements for Multimerization of the Pathogen Receptor Dendritic Cell-specific ICAM3-grabbing Non-integrin (CD209) on the Cell Surface*

The myeloid C-type lectin dendritic cell-specific ICAM3-grabbing non-integrin (DC-SIGN, CD209) recognizes oligosaccharide ligands on clinically relevant pathogens (HIV, Mycobacterium, and Aspergillus). Alternative splicing and genomic polymorphism generate DC-SIGN mRNA variants, which have been detected at sites of pathogen entrance and transmission. We present evidence that DC-SIGN neck variants are expressed on dendritic and myeloid cells at the RNA and protein levels. Structural analysis revealed that multimerization of DC-SIGN within a cellular context depends on the lectin domain and the number and arrangement of the repeats within the neck region, whose glycosylation negatively affects oligomer formation. Naturally occurring DC-SIGN neck variants differ in multimerization competence in the cell membrane, exhibit altered sugar binding ability, and retain pathogen-interacting capacity, implying that pathogen-induced cluster formation predominates over the basal multimerization capability. Analysis of DC-SIGN neck polymorphisms indicated that the number of allelic variants is higher than previously thought and that multimerization of the prototypic molecule is modulated in the presence of allelic variants with a different neck structure. Our results demonstrate that the presence of allelic variants or a high level of expression of neck domain splicing isoforms might influence the presence and stability of DC-SIGN multimers on the cell surface, thus providing a molecular explanation for the correlation between DC-SIGN polymorphisms and altered susceptibility to HIV-1 and other pathogens.

The myeloid C-type lectin dendritic cell-specific ICAM3grabbing non-integrin (DC-SIGN, CD209) recognizes oligosaccharide ligands on clinically relevant pathogens (HIV, Mycobacterium, and Aspergillus). Alternative splicing and genomic polymorphism generate DC-SIGN mRNA variants, which have been detected at sites of pathogen entrance and transmission. We present evidence that DC-SIGN neck variants are expressed on dendritic and myeloid cells at the RNA and protein levels. Structural analysis revealed that multimerization of DC-SIGN within a cellular context depends on the lectin domain and the number and arrangement of the repeats within the neck region, whose glycosylation negatively affects oligomer formation. Naturally occurring DC-SIGN neck variants differ in multimerization competence in the cell membrane, exhibit altered sugar binding ability, and retain pathogen-interacting capacity, implying that pathogen-induced cluster formation predominates over the basal multimerization capability. Analysis of DC-SIGN neck polymorphisms indicated that the number of allelic variants is higher than previously thought and that multimerization of the prototypic molecule is modulated in the presence of allelic variants with a different neck structure. Our results demonstrate that the presence of allelic variants or a high level of expression of neck domain splicing isoforms might influence the presence and stability of DC-SIGN multimers on the cell surface, thus providing a molecular explanation for the correlation between DC-SIGN polymorphisms and altered susceptibility to HIV-1 and other pathogens.
Structurally, DC-SIGN contains a carbohydrate-recognition domain, a neck region composed of eight 23-residue repeats, and a transmembrane region followed by a cytoplasmic tail containing recycling and internalization motifs (5, 30 -32). Analysis of recombinant molecules has revealed that the monomeric lectin domain has low affinity for carbohydrates, whereas full-length DC-SIGN molecules form tetramers through their neck domain, thus allowing high affinity recognition of specific ligands (33)(34)(35). In addition to this prototypical structure, alternative splicing events generate DC-SIGN isoform transcripts whose presence exhibits inter-individual variations (36). The numerous DC-SIGN isoform transcripts reported to date include an alternative cytoplasmic tail, an absent transmembrane region, truncated lectin domains, and a variable number of repeats within the neck domain (36). Moreover, the 23-residue repeat region of DC-SIGN is polymorphic at the genomic level (37,38). Five different alleles for the DC-SIGN neck domain have been identified to date, whose presence correlates with altered susceptibility to HIV-1 transmission (38). The functional relevance of the DC-SIGN neck variants has been further suggested by their detection at mucosal HIV transmission sites (39). Given the involvement of the neck domain in recombinant DC-SIGN multimerization, we hypothesized that the existence of this large array of polymorphic variants might have an impact on the repertoire of pathogen recognition by dendritic cells, as well as on the establishment of interactions between dendritic cells and other cell types. We have characterized naturally occurring alternative splicing isoforms, allelic variants and mutant isoforms of DC-SIGN in terms of surface receptor multimerization and adhesive and pathogen-recognition capabilities, and found that the lectin domain contributes to DC-SIGN multimerization on the cell surface, that glycosylation of the neck domain negatively regulates formation of multimers, and that a neck domain with a single 23-residue repeat is sufficient to mediate DC-SIGN multimerization on the cell surface. Functional comparison of the distinct constructs revealed that the basal multimerization of DC-SIGN does not correlate with enhanced binding to endogenous or pathogenic ligands, indicating that pathogen-induced cluster formation predominates over the basal multimerization capability of the DC-SIGN molecule and is the driving force for the DC-SIGN-dependent pathogen capture and internalization.

Generation of MDDCs
Human peripheral blood mononuclear cells were isolated from buffy coats from healthy donors over a Lymphoprep (Nycomed, Norway) gradient according to standard procedures. Monocytes were purified from peripheral blood mononuclear cells by magnetic cell sorting using CD14 microbeads (Miltenyi Biotech, Bergisch Gladbach, Germany), and immediately subjected to the dendritic cell differentiation protocol (40). Monocytes were cultured for 5-7 days in complete medium with 1000 units/ml granulocyte macrophage-colony stimulating factor (Schering-Plough, Kenilworth, NJ) and 1000 units/ml interleukin-4 (PreProtech, Rocky Hill, NJ) cytokine addition every second day, to obtain a population of immature MDDCs.

Stable and Transient Transfection of DC-SIGN Mutants and Isoforms
For transient transfections, COS-7 or HEK293T cells were transfected with SuperFect (Qiagen) using pCDNA3.1-based expression plasmids containing the distinct isoforms or mutants of the DC-SIGN cDNA. To generate stable transfectants, K562 cells were transfected with pCDNA3.1-based constructs using SuperFect and cultured in complete medium supplemented with 300 g/ml G418 (Invitrogen). Stable DC-SIGN expression of the selected population was verified using the anti-DC-SIGN MR1 monoclonal antibody (13). Isolation of K562-DC-SIGN 1A expressing different levels of DC-SIGN was accomplished by cell sorting after staining with the MR1 monoclonal antibody (13).

Flow Cytometry and Antibodies
Cellular phenotypic analysis was carried out by indirect immunofluorescence, using FITC-labeled goat anti-mouse antibody (Serotec, Oxford, UK). Monoclonal antibodies used for cell surface staining included MR1 (directed against the lectin domain of DC-SIGN), and the supernatant from the mouse myeloma P3-X63Ag8 (X63) was used as the control. All incubations were done in the presence of 50 g/ml human IgG to prevent binding through the Fc portion of the antibodies. Flow cytometry analysis was performed with an EPICS-CS (Coulter Científica, Madrid, Spain) using log amplifiers.

Immunofluorescence
Cells were resuspended in PBS and allowed to adhere onto poly-L-lysine-coated coverslips for 60 min at 37°C. After a brief washing step with PBS, cells were fixed and permeabilized in a 1:1 solution of acetone:methanol for 10 min at Ϫ20°C, washed, and stained with the MR1 monoclonal antibody (13) followed by an incubation with an FITC-labeled goat anti-mouse antibody. Coverslips were mounted in fluorescent mounting medium (DakoCytomation, Carpinteria, CA), and representative fields were photographed through an oil immersion lens on a Nikon Eclipse E800 microscope equipped for epifluorescence or by confocal microscopy.

Generation of Polyclonal Antisera against DC-SIGN Structural Domains
Peptides based on the sequence of the sixth repeated domain of the DC-SIGN neck region (GELPEKSKQQEIYQELTRL-KAAV), and the region between residues 6 and 33 of the cytoplasmic tail (EPRLQQLGLLEEEQLRGLGFRQTRGYKS), were synthesized by the multiple antigen peptide system (13). New Zealand White rabbits were immunized by subcutaneous injec-

Functional Characterization of DC-SIGN Isoforms and Mutants
C. albicans and A. fumigatus Binding Assays-Conidia were labeled with 0.1 mg/ml FITC for 1 h at room temperature and extensively washed. For conidia-binding assays (25), cells were washed, resuspended in complete medium, and pretreated for 20 min at room temperature with anti-DC-SIGN (MR1) or an isotype-matched irrelevant antibody (X63). Then cells were incubated with FITC-labeled A. fumigatus or C. albicans conidia at the indicated ratios for 30 min at room temperature. After extensive washing, cells were fixed with 2% paraformaldehyde in PBS for 1 h at 4°C, washed, and analyzed by flow cytometry.
DC-SIGN-dependent Adhesion Assays-DC-SIGN-dependent adhesion was evaluated using Saccharomyces cerevisiae mannan as specific ligand. 96-well microtiter EIA II-Linbro plates were coated overnight with mannan at 50 g/ml in PBS at 4°C, and the remaining sites were blocked with 0.5% bovine serum albumin for 2 h at 37°C. Cells were labeled in RPMI 0.5% bovine serum albumin with the fluorescent dye 2Ј,7Ј-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein acetoxymethyl ester (Molecular Probes, The Netherlands) at 37°C and then preincubated for 20 min with either the isotype-matched control X63 or the function-blocking MR1 antibodies. Cells were then allowed to adhere to each well for 15 min at 37°C. Unbound cells were removed by three washes with RPMI 0.5% bovine serum albumin, and adherent cells were quantified using a fluorescence analyzer. Where specified, results are presented as "DC-SIGN-dependent binding," defined as: DC-SIGNdependent binding ϭ (% bound cells in the presence of P3X63 Ϫ % bound cells in the presence of MR1).
Leishmania Amastigote Binding Assays-Cells were washed in PBS 1 mM EDTA, resuspended in complete medium and 5,6-carboxyfluorescein succinimidyl ester-labeled parasites were added onto the cells at a 10:1 (amastigotes:cell) ratio, and incubated at room temperature for 30 min. Afterward, cells were fixed and analyzed by flow cytometry. For inhibition assays, cells were preincubated for 10 min at room temperature with either MR1 antibody or an irrelevant antibody (X63) in complete medium before parasite addition.
DC-SIGN Internalization-Cells were washed, resuspended in complete medium, and incubated with MR1 antibody (13) for 1 h at 4°C to prevent DC-SIGN internalization. After extensive washing, cells were placed at 37°C to allow internalization to occur. At the indicated time points, internalization was stopped by adding of cold PBS, and cells were immediately placed at 4°C. To detect the remaining membrane-bound MR1 antibody, an FITC-labeled goat anti-mouse antibody was added, incubated for 30 min at 4°C, and analyzed by flow cytometry. All incubations were done in the presence of 50 g/ml human IgG to prevent binding through the Fc portion of the antibodies.
Ebola GP1-Fc Binding Assays-Cells were washed in PBS and 1 mM EDTA, resuspended in complete medium, and incubated with GP1-Fc either in the presence of a monoclonal antibody against DC-SIGN (MR1) or an irrelevant antibody (X63) for 20 min at 4°C. Then, cells were incubated with a phycoerythrinlabeled polyclonal antiserum against human IgG Fc (Beckman Coulter), and analyzed by flow cytometry.
Sugar-coated Fluorescent Bead Binding to DC-SIGN-Synthetic fluorescein-labeled fucose-or Lewis x -containing polyacrylamide beads (FITC-PAA-NAc-Gal, FITC-PAA-Fuc, and FITC-PAA-Le x ) were obtained from Lectinity (Moscow, Russia). After washing with PBS and 1 mM EDTA, transiently transfected HEK293T cells were resuspended in complete medium, and sugar-PAA-FLU beads were added to a final concentration of 20 g/ml and incubated at 37°C for 30 min. After extensive washing, cells were fixed for 1 h at room temperature, and analyzed by flow cytometry. For inhibition assays, cells were preincubated for 10 min at room temperature with either MR1 antibody or an irrelevant antibody (X63) in complete medium before beads addition. Results from binding assays were expressed as "Binding Index," which represents the DC-SIGNdependent binding relative to DC-SIGN expression levels according to the formula: Binding index ϭ (mean fluorescent intensity (MFI) of cells plus beads Ϫ MFI of cells plus beads in the presence of MR1)/(MFI after MR1 staining/MFI after staining with X63).

NMR Experiments
Binding of soluble glucomannan from Candida utilis (IF) to DC-SIGN transfectants was done by basic Saturation Transfer Difference, as previously described (43).

Western Blot
Total cell lysates were obtained in Nonidet P-40 lysis buffer, and 10 g of each lysate was subjected to SDS-PAGE under reducing or non-reducing conditions and transferred onto an Immobilon polyvinylidene difluoride membrane (Millipore, Bedford, MA). After blocking of the unoccupied sites with 5% nonfat dry milk in 50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.1% Tween 20, protein detection was performed using the Super-Signal West Pico chemiluminescent system (Pierce). Detection of DC-SIGN was carried out using polyclonal antiserum against the C-terminal 20-residue peptide of DC-SIGN (C-20, sc-11038, Santa Cruz Biotechnology, Santa Cruz, CA), amino
SDS-eluted and non-bound materials were resolved by SDS-PAGE and DC-SIGN detection accomplished with specific polyclonal antibodies.

Range of DC-SIGN Alternatively Spliced Isoforms-MDDCs
express a high number of alternatively spliced DC-SIGN mRNA species (36), which are also found at mucosal HIV transmission sites (39). To determine the range of DC-SIGN mRNA species found in MDDC from a single donor, three different set of primers were designed to specifically amplify the prototypical DC-SIGN mRNA (DC-SIGN 1A), or species encoding either an alternative cytoplasmic domain (DC-SIGN 1B) or lacking the transmembrane domain (DC-SIGN ⌬TM) (Fig. 1A). Sequencing of the amplified fragments resulted in the identification of DC-SIGN mRNA species encoding for variants with  FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 alternative cytoplasmic tails and potentially soluble isoforms, with each group including transcripts differing in the neck domain or the carbohydrate-binding region (Fig. 1B). Therefore, and in agreement with previous reports (14,36,39), the DC-SIGN gene gives rise to a large number of alternatively spliced mRNA species, most of which differ in the number of 23-residue repeats within the neck domain, previously demonstrated to mediate multimerization of recombinant DC-SIGN (34).

Expression and Function of DC-SIGN Variants
Next we generated polyclonal antisera specific for either the neck domain (DSG-1) or the prototypic cytoplasmic tail of DC-SIGN 1A (DSG-2). Both DSG-1 and DSG-2 specifically detected the 44-kDa band of the prototypic full-length isoform DC-SIGN 1A, as well as a deletion mutant lacking the lectin domain (8d⌬L), whereas a polyclonal antiserum against the 20 C-terminal residues of the lectin domain (C-20) only detected the full-length molecule (Fig. 1C).
To determine the degree of DC-SIGN multimerization on MDDC, cell surface proteins were biotin-labeled, and streptavidin pulled-down material was analyzed for the presence of DC-SIGN. Under non-reducing conditions, the DSG-2 antiserum detected distinct several bands corresponding to DC-SIGN monomers, dimers, trimers, tetramers, and high order multimers either in the whole extracts and the pull-down (Fig.  1D, left panel, lanes WE and SP, respectively), which suggests that DC-SIGN multimers are found on the cell surface of MDDCs. Analysis of cell surface DC-SIGN molecules from MDDC under reducing conditions also revealed the presence of additional higher mobility bands that were also recognized by the DSG-2 antiserum (Fig. 1D, right panel, lane SP). The same pattern was detected in total lysates of dendritic-like THP-1 cells (14) using a polyclonal antiserum against the whole neck region of the molecule (Fig. 1E), and similar bands could be detected in MDDC lysates with both DSG-1 and DSG-2 antisera (Fig. 1F). Therefore, DC-SIGN isoforms can be detected on the cell surface of monocyte-derived dendritic cells, although to a lower extent than the full-length DC-SIGN 1A.
Contribution of the Lectin Domain to DC-SIGN Multimerization on the Cell Membrane-DC-SIGN multimer formation in MDDC could be readily identified by SDS-PAGE (Fig. 1D) (33,44). In fact, although treatment with the membrane-impermeable cross-linker DTSSP enhanced the formation of high order multimers, DC-SIGN monomers, dimers, and multimers were readily detected by C-20, DSG-1, and DSG-2 antisera under non-reducing conditions (Fig. 1F). The detection of DC-SIGN multimers was almost completely prevented in the presence of reducing agents (Fig. 1, D and F), indicating that disulfide bridges contribute to multimerization. Because all Cys residues within the lectin domain are engaged in intramolecular disulfide bridges (27), we determined the effect of mutating Cys 37 , the only DC-SIGN cysteine residue outside of the lectin domain and located within the cytoplasmic tail. Mutation of Cys 37 had no effect on the degree of formation of DC-SIGN multimers (Fig. 2B, left panel), suggesting that multimerization could be dependent on cysteine residues within the lectin domain. The lectin domain was then removed from either the DC-SIGN prototypic isoform (8d⌬L) or from an isoform with only six repeats (6d⌬L) (Fig. 2A). Both 8d⌬L and 6d⌬L constructs displayed greatly reduced multimerization ability (Fig. 2B, right panel), which indicates that, although DC-SIGN multimerization might be mediated by the neck region (34,35,45), it requires or is stabilized by the lectin domain of the molecule.
Along this line, the presence of lectin domain-lacking constructs (8d⌬L, 7d⌬L, and 6d⌬L) had a negative impact on the degree of multimerization of DC-SIGN 1A, as we observed a lower level of DC-SIGN 1A multimers in the presence of these deletion constructs (Fig. 2C). This could be explained by an increased formation of heteromultimers (formed by DC-SIGN 1A and constructs lacking the lectin domain), which might exhibit lower stability in the presence of denaturing detergent, thus precluding its detection. If so, the existence of heteromultimers could be demonstrated by coprecipitation experiments on lysates from cells cotransfected with DC-SIGN 1A and 8d⌬L. The fact that the 8d⌬L isoform was pulled down after immunoprecipitation of lectin domain-containing molecules with the MR1 monoclonal antibody (Fig. 2D) confirms that lectin domain-lacking constructs associate with the prototypic DC-SIGN 1A isoform, suggests that heteromultimers of DC-SIGN 1A and 8d⌬L are more sensitive to the presence of denaturing agents than DC-SIGN 1A homomultimers and confirms a role for the lectin domain of DC-SIGN in the formation of stable oligomers.
Structural Requirements of the Neck Domain for DC-SIGN Multimerization-Although the neck domain is absolutely required for the formation of multimers of recombinant nonglycosylated DC-SIGN and DC-SIGNR (34,35,45), its role in DC-SIGN multimerization on the cell membrane remains unclear. To address this issue, we analyzed the pattern of multimerization of the prototypic full-length molecule (1A), naturally occurring (4d, 4dЈ, 2d, and 1d) or in vitro generated (3d) isoforms differing in the number and order of the neck region repeats, and constructs mutated at the N-linked glycosylation site (1AN/Q and 1dN/Q) (Fig. 3A). Transient transfection revealed that the distinct DC-SIGN isoforms differed in their ability to form oligomers. A high proportion of full-length DC-SIGN 1A appeared as multimers, whereas deletion of half the neck region (4d) resulted in a considerable reduction of high order multimers (Fig. 3A). By contrast, isoforms 3d and 2d, whose neck regions are composed of three and two repeats, exhibited an oligomerization ability roughly similar to that of the full-length molecule, whereas isoform 1d showed the weakest oligomerization (Fig. 3A). These results indicate that the presence of at least two repeats within the neck region is sufficient for DC-SIGN multimerization. On the other hand, the lower multimerization of 4d suggests that there is no direct correlation between the length of the neck region and oligomerization, and that the distinct repeats within the neck region might not be functionally equivalent. This hypothesis was confirmed when comparing the low multimerization capability of 4d (composed of neck repeats 1, 6, 7, and 8) with the normal (similar to 1A) oligomerization pattern of 4dЈ isoform, whose neck region is composed of repeats 1, 2, 3, and 8 (Fig. 3A), thus confirming that multimerization capability of DC-SIGN on the cell membrane is dependent not only on the number of neck repeats but also on their arrangement, and that the repeats within the neck region of DC-SIGN are not functionally interchangeable.
In agreement with the results obtained after transient transfection, the 4d isoform also exhibited a greatly reduced proportion of DC-SIGN multimers when stably expressed in K562 cells, whereas multimerization of 2d isoform was similar to that of DC-SIGN 1A (Fig. 3B), a finding also observed after transfection in T lymphoblastoid Jurkat cell (data not shown). Functional analysis of the three isoforms in K562 transfectants revealed that 1A, 4d, and 2d bound soluble C. utilis glucomannan (IF), as determined by one-dimensional saturation transfer difference (43,46) (Fig. 3C), and were internalized after MR1mediated engagement (Fig. 3D). Therefore, it can be concluded that the degree of multimerization of functional DC-SIGN isoforms on the cell surface is cell-type independent and does not influence the ligand-induced internalization of the molecule.
The inability of the first repeat to mediate multimerization (1d in Fig. 3A), and the fact that it contains the only potential N-glycosylation site of DC-SIGN, prompted us to determine the contribution of glycosylation to DC-SIGN oligomerization. Replacement of Asn 80 for Gln in the context of the full-length molecule (1AN/Q) greatly increased the proportion of DC-SIGN multimers (compare 1A and 1AN/Q in Fig. 3A) and suggests that glycosylation of the first neck repeat negatively affects DC-SIGN multimerization. The negative influence of glycosylation on multimerization was even more evident upon analysis of the 1dN/Q mutant, whose neck domain is formed only by the first repeat with the Asn 80 /Gln replacement. Unlike the 1d isoform, oligomers (and even high order multimers) of the 1dN/Q mutant could be easily detected (Fig. 3A). In fact, and like in the case of 1AN/Q, no 1dN/Q monomers were observed under non-reducing conditions (Fig. 3A). Therefore, FIGURE 2. Determination of structural requirements for DC-SIGN multimerization. A, schematic representation of the DC-SIGN alternatively spliced isoforms (1A, 4d, 4dЈ, 3d, 2d, and 1d), mutants (1AC/S, 1AN/Q, and 1dN/Q) and chimeric molecules (8d⌬L, 7d⌬L, and 6d⌬L) used throughout the study. The Cys 37 residue is indicated by a cross ( †) on the transmembrane region, and the presence of potential N-glycosylation sequence is indicated by a black dot on the first repeat of the neck domain. B and C, COS-7 cells transiently transfected with the indicated DC-SIGN constructs were lysed, resolved by SDS-PAGE, and subjected to Western blot using DSG-2 (B), or DSG-1 and DSG-4 (C) polyclonal antisera. D, lysates from COS-7 cells transiently transfected with the indicated DC-SIGN constructs were immunoprecipitated with a monoclonal antibody against the DC-SIGN lectin domain (MR1) or a control antibody (CNT), and immunoprecipitated proteins were resolved by SDS-PAGE and subjected to Western blot using DSG-1 polyclonal antiserum. Immunoprecipitated proteins were analyzed in parallel with whole cell lysates (WE). FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 glycosylation of the first repeat in the neck region impairs multimerization of DC-SIGN molecules.

Expression and Function of DC-SIGN Variants
Influence of Multimerization on DC-SIGN Pathogen Recognition-Despite the differences in their ability to multimerize, transient transfection of the whole range of constructs previously assayed revealed that all of them are capable of binding Candida yeasts and Leishmania amastigotes to a similar extent (supplemental Fig. S1A). To rule out the subtle differences in pathogen binding among the distinct constructs we evaluate Candida and L. pifanoi amastigotes binding by cells expressing decreasing levels of three naturally occurring isoforms (1A, 4d, and 2d), and no significant difference was observed when comparing binding by cells expressing similar levels of the three constructs (supplemental Fig. S1B and not shown). However, the results showed that the binding ability of the distinct DC-SIGN isoforms correlate with their expression level (supplemental Fig. S1B). Therefore, because the multimerization ability of the 4d isoform is considerably lower than that of 1A and 2d (see Fig. 3A), these results indicate that the recognition of C. albicans yeasts or L. infantum amastigotes by distinct DC-SIGN isoform/mutants does not correlate with their multimerization degree. Consequently, the multimerization degree of an isoform does not predict its pathogen-binding ability.
Influence of Multimerization on Sugar Recognition by DC-SIGN-The lack of correlation between multimerization degree and pathogen-binding ability of DC-SIGN isoforms could be explained by the large amount of DC-SIGN ligands immobilized on the pathogen surface, which would drive the formation of DC-SIGN-containing clusters (47,48) and might obscure the contribution of the affinity/avidity of individual molecules/oligomers to the whole interaction. To avoid pathogen-induced clustering effects on the membrane, we assess the ability of the distinct DC-SIGN constructs to be retained by sugars after membrane solubilization. As shown in Fig. 4A (upper panels), except 1d, all DC-SIGN constructs were specifically retained by mannan (a polysaccharide that blocks most DC-SIGN interaction). However, analysis of molecules not retained by mannan (supernatant) revealed that constructs 1A, 1AN/Q, 4d, and 1dN/Q are retained with higher efficiency than the 4dЈ, 3d, and 2d constructs (Fig. 4A, lower panels). Monomers were preferentially retained by mannan within the strong mannan-binding and N-glycosylation-containing constructs (1A and 4d) (lanes 1A and 4d in the left panels of Fig. 4A). By contrast, those exhibiting lower binding to mannan (4dЈ, 3d, and 2d) were preferentially retained as multimers, as monomers were almost exclusively detected in the supernatant (lanes 4dЈ, 3d, and 2d in left panels of Fig. 4A). Furthermore, and in agreement with the negative effect of N-glycosylation on DC-SIGN multimerization, the 1AN/Q and 1dN/Q constructs were preferentially retained as multimers. Therefore, functional analysis of detergent-solubilized cellular DC-SIGN demonstrates that multimer formation compensates for the lower mannan-binding affinity of certain DC-SIGN constructs after membrane solubilization, an effect that becomes even more evident when less-than-optimal sugar ligands (NAc-Gal) were used, which only retained lectin multimers (Fig. 4B). Therefore, this set of data indicates that the number and arrangement of the repeats within the neck domain directly influences the specificity and the sugarbinding ability of the DC-SIGN lectin domain.
To further evaluate the relevance of DC-SIGN cell surface multimerization on ligand binding, cell surface expressed 1d and 1dN/Q constructs were compared in their ability to bind FITC-PAA-Fucose and Lewis x beads. 1dN/Q, which appears almost exclusively as multimers, displayed a stronger beadbinding activity than 1d, whose multimers can barely be detected, and the same finding was observed at three distinct cell surface expression levels (Fig. 4C). These results further support the involvement of cell surface DC-SIGN multimerization in ligand binding, and establish N-linked-glycosylation as a critical parameter for the DC-SIGN ligand-binding activity on the cell surface.  FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7

Structural and Functional Characterization of Polymorphic
Variants of DC-SIGN-The above results demonstrate that the neck region is an important determinant in the ligand-binding activity of DC-SIGN on the cell surface. It has been reported that, among the polymorphisms in the DC-SIGN gene (38, 49 -52), those affecting the length of the neck domain correlate with altered susceptibility to HIV-1 infection (38). In fact, similar findings have been reported in the case of the related DC-SIGNR and susceptibility to HIV-1 and severe acute respiratory syndrome infection (53)(54)(55). To evaluate the functional significance of polymorphic DC-SIGN neck domains, three distinct allelic variants, whose neck domains contain only seven repeats, were identified at the genomic DNA and RNA level (Fig. 5, A  and B) and functionally characterized. These polymorphic variants lack repeats 3, 5, or 7, but their multimerization ability (Fig.  5B), cell surface expression (Fig. 5C), and ligand-induced internalization capability (Fig. 5D) were found to be indistinguishable from that of the prototypic molecule. Moreover, the three variants displayed unaltered capacity for recognition of Leish-mania and Aspergillus (supplemental Fig. S2A) and mediated cellular binding to Ebola GP1-Fc and Mannan (supplemental Fig. S2B) and were retained by agarose-bound mannan after membrane solubilization (supplemental Fig. S2C), and mediated binding of C. utilis glucomannan to cells as determined by one-dimensional saturation transfer difference NMR (supplemental Fig. S2D). Therefore, DC-SIGN polymorphic variants lacking a single neck domain repeat (3, 5, or 7) exhibit functional activities that are similar to those exhibited by the prototypic DC-SIGN molecule.
Because altered susceptibility to infections has been mostly observed in individuals with heterozygosity at the DC-SIGN (or DC-SIGNR) gene (38,(53)(54)(55), we next evaluated the influence of DC-SIGN polymorphic variants (-D3, -D5, and -D7, Fig. 5B) on the expression, multimerization, and functional capability of the prototypic molecule when expressed on the same cell. Transient transfection experiments demonstrated that the expression of any of the polymorphic variants had no influence on the DC-SIGN 1A total or cell surface expression (Fig. 6, A and B). Like in the case of the 1A/8d⌬L cotransfectants, heterooligomers might have an increased sensitivity to denaturing agents. However, coimmunoprecipitation experiments with epitope-tagged molecules demonstrated that the DC-SIGN 1A molecules preferentially formed homo-oligomers, and associate weakly to seven repeat-containing polymorphic variants (Ͻ1% of the prototypic DC-SIGN molecules are engaged in hetero-oligomer formation) (Fig. 6C). Therefore, the shorter polymorphic variants can be expressed on the cell surface but tend to form homo-oligomers and associate very weakly with the prototypic DC-SIGN 1A full-length isoform. This result would imply that cells heterozygous at the DC-SIGN gene might almost exclusively express homo-oligomers on the cell surface. We tested this hypothesis by analyzing the DC-SIGN isoform expression and multimerization state in MDDC from a donor previously identified as a CD209 heterozygote (see Fig.  5A). The prototypic and shorter variant of DC-SIGN were expressed to a similar extent in heterozygous dendritic cells, but no evidence was found of hetero-oligomer formation (Fig.   6D), confirming that DC-SIGN multimerization takes place preferentially among variants whose neck region has identical structure.
The low percentage (or impaired stability) of lectin heterooligomers might explain the reduced pathogen-binding capacity exhibited by cells coexpressing allelic variants of DC-SIGNR (53) and the correlation between DC-SIGN/DC-SIGNR neck region heterozygosity and susceptibility to viral infection (38,(53)(54)(55). Consequently, DC-SIGN-dependent activities of cells coexpressing different DC-SIGN allelic variants were evaluated on transiently transfected cells. As shown in Fig. 6B, coexpression of the DC-SIGN-D7 isoform (which lacks the seventh neck domain repeat) did not significantly affect the ability of the prototypic DC-SIGN 1A isoform to bind immobilized mannan. Along the same line, capture of fucose-or Lewis x -coated polyacrylamide beads by DC-SIGN 1A was not affected by the coexpression of the DC-SIGN-D3 isoform (which lacks the third neck domain repeat) ( Fig. 7A and not shown). These results indicated that expression of polymorphic variants with shorter  FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 neck domains does not significantly alter the pathogen recognition ability of cells expressing the prototypic DC-SIGN 1A isoform.

Expression and Function of DC-SIGN Variants
Finally, because sugar precipitation had previously allowed the identification of functional differences among alternatively spliced isoforms (Fig. 4), lysates from MDDCs coexpressing DC-SIGN 1A and DC-SIGN-D7 were subjected to precipitation with mannan-agarose. Whereas a single band (corresponding to DC-SIGN 1A) was specifically retained from the 1A/1A dendritic cells, both DC-SIGN 1A and DC-SIGN-D7 isoforms were equally retained by mannan-agarose when the dendritic cell lysate from the 1A/-D7 donor was used (Fig. 7B). Therefore, DC-SIGN 1A and DC-SIGN-D7 are retained by mannan to a similar extent, confirming that shorter neck polymorphic variants of DC-SIGN retain their sugar-recognition ability, showing no differences from that of the prototypic fulllength DC-SIGN 1A molecule.

DISCUSSION
DC-SIGN-dependent binding and uptake of clinically relevant pathogens by dendritic cells relies on the lectin ability to bind mannose-and fucose-containing glycans (56). Studies on recombinant molecules have demonstrated that the avidity of such interactions is mediated through multimerization of the lectin, which is accomplished through intermolecular associations mediated by the neck domain of the molecule (27,34). The neck region of DC-SIGN is composed of eight 23-amino acid repeats, which are encoded in a single exon whose polymorphism has been already demonstrated (38,51). In fact, DC-SIGN alleles with 4 -9 repeats within the neck region-coding exon have been described (51), and heterozygosity at this specific exon correlates with altered susceptibility to HIV-1 infection (38). Besides, numerous DC-SIGN alternatively spliced isoforms have been described at the mRNA level (14,36). The combination of alternative splicing and genomic polymorphism predicts that a large repertoire of DC-SIGN protein isoforms might exist, most of which would differ in the size of the neck domain (14,36,38,51). However, to date, the functional characterization of DC-SIGN isoforms and allelic variants on the cell membrane had not been addressed. In the present manuscript we present evidences that 1) DC-SIGN alternatively spliced mRNA species give rise to proteins that are expressed at the cell membrane on monocyte-derived dendritic cells, cell lines, and transfectants; 2) DC-SIGN alternatively spliced isoforms differ in their multimerization capability and sugar-binding ability; 3) the presence of two repeats within the neck domain is sufficient for DC-SIGN multimerization; 4) the neck domain repeats are not functionally interchangeably, because the number and arrangement of repeats within the neck domain critically determines the multimerization and ligand-binding ability; 5) the lectin domain of DC-SIGN stabilizes or contributes to the neck region-dependent multimerization of DC-SIGN, which is negatively influenced by the N-linked glycosylation of the first neck domain repeat; 6) basal multimerization of the molecule does not predict the pathogen-binding ability and does not correlate with ligandinduced internalization; and 7) polymorphic variants differing in neck domain composition can self-associate, but multimerize very poorly with the prototypic full-length molecule, suggesting that the DC-SIGN molecules on the cell surface predominantly appear as homo-multimers. The data here presented constitutes the first demonstration that alternative splicing and polymorphic variants of DC-SIGN are expressed on monocyte-derived dendritic cells, where they exhibit altered multimerization and carbohydratebinding abilities (splicing variants) and tend to segregate from the prototypic molecule forming homo-multimers (polymorphic variants with a shorter neck domain).
Whether DC-SIGN and related lectins are bona fide pathogen-recognition receptors or antigen-binding receptors whose function is subverted by pathogens is still unclear (4,57). Our results indicate that the various alternatively spliced isoforms differ in their ability to be retained by immobilized mannan, whereas all of them are equally efficient in terms of pathogen binding. We hypothesize that the large amount of DC-SIGN ligands on the surface of interacting pathogens compensate for the distinct affinity and multimerization ability of the isoforms. If this is the case, pathogen-induced formation of DC-SIGNcontaining clusters on the cell surface would counterbalance for the diminished multimerization ability of certain isoforms and would justify the large range of pathogens bound and internalized via DC-SIGN. Therefore, according to this hypothesis, isoforms would have a physiological role (increasing the range of soluble antigens bound and internalized by DC-SIGN), but would not have a major impact on the range of pathogens bound by DC-SIGN. Further studies are needed to clarify these issues, because it is currently unknown whether the basal multimerization of DC-SIGN on the cell surface (33,44) is exclusively mediated by intermolecular interactions or is a soluble ligand-induced event. In this regard, all the experiments performed in the present study were done after extensive washing of the cells with EDTA, to prevent any carbohydrate-DC-SIGN interaction that might affect multimerization of the molecule on the cell surface.
Sequence analysis has allowed the definition of 23-residue repeats within the neck region of DC-SIGN, which is sometimes divided into 7.5 repeats to account for the presence of an unrelated and unique sequence at the N-terminal half of the first repeat (34). Ultracentrifugation and cross-linking of recombinant truncated DC-SIGN molecules have established that removal of the two N-terminal repeats only partially affected the tetramerization ability, whereas recombinant proteins containing only repeats 7-8 formed partially dissociating dimers. This has led to the proposal that repeats close to the lectin domain mediate dimer formation while the membrane proximal repeats are required for tetramer formation (34). Our results with transient and stable transfectants of the naturally occurring DC-SIGN 4d and 2d isoforms, which include repeats 1, 6, 7, and 8 and 1 and 2 (Fig. 3A), indicate that the two more N-terminal domains are sufficient for multimerization in a cellular context, a fact further confirmed by the very different multimerization capability of the 4d (1, 6, 7, and 8) and 4dЈ (1, 2, 3, and 8) isoforms. The importance of repeats 1 and 2 for the ability of DC-SIGN to multimerize in the cell membrane is even more evident when considering that the DC-SIGN 1d isoform (containing only repeat 1) does not multimerize, and that removal of the N-glycosylation site (1dN/Q mutant) allows multimerization within a cellular context. In addition, the 3d construct, which includes the first repeat followed by the N-terminal half of repeat 2, the C-terminal half of repeat 7 and the entire repeat 8, also exhibits an efficient multimerization capability within a cellular context. Therefore, essential residues for multimerization can be mapped to the sequence GELSE at the beginning of the second repeat, which includes a serine residue unique among the repeats and contributes to the multimerization ability of recombinant DC-SIGNR (58). These results demonstrate the critical role of repeats 1 and 2 for DC-SIGN multimerization, because repeat 1 is capable of mediating multimer formation, and the mere presence of repeat 2 appears sufficient to overcome the inhibitory effect of the N-glycosylation at repeat 1. These results are compatible and extend previous data on the multimerization capability of recombinant DC-SIGN/ DC-SIGNR molecules, and establish neck glycosylation as an important parameter to limit the degree of DC-SIGN multimerization in the cell.
The combination of genomic polymorphism and alternative splicing at the DC-SIGN gene results in the generation of a large number of isoforms/allelic variants of the molecule. Considering their variable multimerization capability, and the higher avidity displayed by multimers, it is tempting to speculate that the existence of all these variants might endow macrophages and dendritic cells with a broader repertoire of ligand-binding affinity and/or specificity. In fact, the ability of mannan-agarose to differentially retain the various DC-SIGN splicing isoforms (Fig. 4) would support this hypothesis. On the other hand, an alternative function for the numerous DC-SIGN isoforms could be the modulation of full-length DC-SIGN-dependent functions. In this regard, and like the lectin domain-lacking chimeric constructs (Fig. 2), isoforms with truncated lectin domains might reduce the effective concentration of full-length DC-SIGN molecules on the cell membrane, thus impairing its multimerization on the cell surface and, consequently, the binding and uptake of pathogens/ligands containing limiting amounts of sugar ligands.
Regarding polymorphic variants, our results indicate that the number of DC-SIGN allelic variants is greater than previously thought. The study of Barreiro and Liu (38,51) has defined polymorphisms within the neck region of DC-SIGN and classified them according to the number of repeats. However, and at least within the Spanish population, the allelic variants containing only 7 neck repeats are not structurally identical, and three distinct alleles have been identified which differ in the missing repeat within the neck domain (D3, D5, and D7). Therefore, it is likely that most of the previously defined CD209 alleles are really heterogeneous in terms of the arrangement of the neck domain repeats they contain. On the other hand, and despite the association found between neck domain heterozygosity at the CD209 and CD209L genes and altered susceptibility to HIV-1 (38,55), hepatitis C (59, 60), or severe acute respiratory syndrome infection (53), the polymorphic variants that we have characterized exhibit similar homo-multimerization capability and pathogen-and carbohydrate-binding specificity as the fulllength molecule. However, the polymorphic variants containing seven repeats (-D3, -D5, and -D7) exhibit a very weak ability to assemble into hetero-multimers with the full-length DC-SIGN 1A prototypic molecule, as hetero-multimers cannot be observed by Western blot and an extremely low percentage of the -D3 variant can be coprecipitated with DC-SIGN 1A (Fig.  6). This result is in contrast to the reported ability of recombinant polymorphic forms of DC-SIGNR to engage in stable homo-and hetero-tetramers (58). However, we feel that this is only an apparent discrepancy, because the N-linked glycosyla- FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 tion of the N-terminal neck repeat limits the extent of multimerization of the molecular within a cellular context (Fig. 2) and, therefore, recombinant molecules (which are devoid of glycosylation) might display an enhanced tendency to multimerize. Whether the reduced ability of DC-SIGN polymorphic variants to associate with the full-length molecule contributes to the altered susceptibility of heterozygous individuals to various infections remains to be determined. However, the preferential formation of homo-multimers in heterozygous individuals must lead to a reduction (50%) in the number of multimers containing the full-length DC-SIGN 1A molecule, what might affect the recognition of pathogens with a limiting amount of carbohydrate ligands. The fact that CD209 gene promoter polymorphisms, thought to affect DC-SIGN cell surface levels, also associate with altered susceptibility to HIV-1 (52), Dengue (37), and tuberculosis (49) is compatible with the above explanation. Consequently, although further studies are required, our results demonstrate that expression of neck domain splicing and allelic variants influence the presence and stability of DC-SIGN multimers on the cell surface, and provide relevant clues about the underlying molecular mechanisms for the association between DC-SIGN polymorphisms and altered susceptibility to clinically relevant pathogens.