Structural insight into the recognition of pathogen-derived phosphoglycolipids by C-type lectin receptor DCAR

The C-type lectin receptors (CLRs) form a family of pattern recognition receptors that recognize numerous pathogens, such as bacteria and fungi, and trigger innate immune responses. The extracellular carbohydrate-recognition domain (CRD) of CLRs forms a globular structure that can coordinate a Ca2+ ion, allowing receptor interactions with sugar-containing ligands. Although well-conserved, the CRD fold can also display differences that directly affect the specificity of the receptors for their ligands. Here, we report crystal structures at 1.8–2.3 Å resolutions of the CRD of murine dendritic cell-immunoactivating receptor (DCAR, or Clec4b1), the CLR that binds phosphoglycolipids such as acylated phosphatidyl-myo-inositol mannosides (AcPIMs) of mycobacteria. Using mutagenesis analysis, we identified critical residues, Ala136 and Gln198, on the surface surrounding the ligand-binding site of DCAR, as well as an atypical Ca2+-binding motif (Glu-Pro-Ser/EPS168–170). By chemically synthesizing a water-soluble ligand analog, inositol-monophosphate dimannose (IPM2), we confirmed the direct interaction of DCAR with the polar moiety of AcPIMs by biolayer interferometry and co-crystallization approaches. We also observed a hydrophobic groove extending from the ligand-binding site that is in a suitable position to interact with the lipid portion of whole AcPIMs. These results suggest that the hydroxyl group-binding ability and hydrophobic groove of DCAR mediate its specific binding to pathogen-derived phosphoglycolipids such as mycobacterial AcPIMs.

The recognition of pathogens by the innate immune system is one of the initial steps leading to a robust and well-orchestrated immune response (1). This physiological process is in part made possible by the presence of proteins at the surface of the immune cell called pattern-recognition receptors. Among these, C-type lectin receptors (CLRs) 3 form a family of pattern recognition receptors that were shown to recognize not only a broad spectrum of pathogens, but also endogenous molecules released by stressed or dying cells, through their interaction with ligands of various types and structures (2)(3)(4). Most of these receptors share the same domain organization: they are type II transmembrane proteins with a cytoplasmic intracellular tail, a transmembrane domain, and an extracellular domain composed of both the stalk region as well as the compact and globular carbohydrate-recognition domain (CRD) (5). Through its ability to coordinate Ca 2ϩ ions, the CRD is, for most CLRs, of paramount importance to bind sugar-containing ligands.
Dendritic cell immunoactivating receptor (DCAR, or Clec4b1) is a CLR expressed by myeloid cells that was originally described by high sequence similarity with dendritic cell immunoreceptor 1 (DCIR1, or Clec4a2) in mouse (6). Although they share 91% similarity in their CRDs, differences were observed in their stalk and intracellular domains. Contrarily to DCIR1, which has an immunoreceptor tyrosine-based inhibitory motif (ITIM) within its own intracellular domain, the adaptor mole-cule Fc receptor ␥ chain (FcR␥) harboring an immunoreceptor tyrosine-based activation motif (ITAM) is required for DCARdependent signaling (6,7).
We recently reported that DCAR functions as a receptor for mycobacteria through the recognition of unique phosphoglycolipids abundantly found in mycobacterial envelopes, called acylated phosphatidyl-myo-inositol mannosides (AcPIMs) (8). This induces MCP-1 production from monocyte-derived inflammatory cells and promotes a Th1-mediated protective response, thereby contributing to the immune response against mycobacteria in mouse (7). However, the molecular and structural basis of phosphate-containing glycolipid recognition by DCAR remains to be clarified. Human DCIR and blood dendritic cell antigen 2 (BDCA-2) are closely related to DCAR, but its own human ortholog has yet to be identified (9,10). However, AcPIMs induce similar sets of inflammatory genes in human dendritic cells (7), implying that humans express unidentified AcPIM receptor, and thus the determination of murine DCAR structure may provide valuable information for the identification of novel bacterial receptor(s).
The present study shows that DCAR has an overall typical C-type lectin fold that allows it to bind carbohydrates in similar fashion to other CLRs. Structural and mutagenesis analysis also revealed its characteristics for the suitable recognition of unique mycobacterial phosphoglycolipids.

DCAR preparation, crystallization, and structural determination
The CRD (residues 78 -209) of murine DCAR was expressed as inclusion bodies in Escherichia coli and then refolded in vitro by a dilution method. Refolding of DCAR did not require Ca 2ϩ ions, and refolded proteins were purified by anion exchange followed by gel-filtration chromatography (Fig. S1, A and B). The elution profile of the refolded CRD displayed a single homogenous and sharp peak at the expected molecular weight, indicating that conformational variations were unlikely and that the CRD behaves as a monomer after refolding (Fig. S1A). DCAR CRD crystals were obtained by the sitting-drop method, the structure was solved at a resolution of 2.3 Å, and the asymmetric unit contained two DCAR molecules ( Table 1). The structure of DCAR is composed of two flanking antiparallel ␣-helices (␣1 and ␣2) and nine ␤-strands (␤1 to ␤9) and is stabilized by three disulfide bonds similar to other CLR family members (Fig. 1). These results thus show that DCAR has an overall typical C-type lectin domain fold.

Structural comparison of DCAR and Mincle
The CRD of the similar CLR Mincle (Fig. 2B), which has been determined as the disaccharide trehalose complex, has 43% sequence similarity with the CRD of DCAR ( Fig. 2A) and was used for structural comparison (11). Superimposition of DCAR and cow Mincle ( Fig. 2B; PDB code 4ZRW) shows that the position of backbone C␣ atoms in both CRDs are similar, with a root-mean-square deviation of 0.627 Å for 129 structurally equivalent C␣ atoms (Fig. 2C). Although Mincle binds three ions, whereas DCAR binds only one, the position of the Ca 2ϩ ion of the primary sugar-binding site is also nearly the same between the two proteins ( Fig. 2, A-C). The canonical EPN (Glu-Pro-Asn) motif of the primary sugar-binding site of Mincle (residues 168 -170) is typically found in glucose/mannose binding CLRs, as opposed to the QPD (Gln-Pro-Asp) motif, which generally indicates a preference toward galactosecontaining ligands (5,12). DCAR bears a similar but unusual EPS (Glu-Pro-Ser) sequence (residues 168 -170), also found in several members of the DCIR family (13). As shown by the crystal structure of hDCIR (PDB code 5B1X), the EPS sequence can successfully coordinate a Ca 2ϩ ion and contribute to mannose recognition (13). In line with this report, DCAR also bound a Ca 2ϩ ion through its EPS sequence, as well as two additional residues of the WND motif (residues 189 -191) which is well-conserved among CLRs (Figs. 1, green filled circles, and 2D). Despite this sequence difference, the Ca 2ϩ -binding site location and the position of the amino acids of DCAR and Mincle are nearly identical (Fig. 2, D-F), with the exception of the residues following the EPN motif in Mincle that interact with another Ca 2ϩ ion and are protruding compared with the corresponding amino acids of DCAR (Fig. 2C). The overall structural comparison therefore confirms that the CRD of DCAR has a typical C-type lectin domain fold and can coordinate a Ca 2ϩ ion through its EPS sequence.

The EPS motif of DCAR is required for optimal function
As mentioned above, DCAR coordinates its Ca 2ϩ ion through an EPS sequence, which was shown to mediate man- Crystal structure of C-type lectin receptor DCAR nose binding by hDCIR (13). This sequence is also found in two other members of the DCIR family. We thus sought to test whether this unusual EPS sequence is necessary for ligand binding by DCAR. For comparison, we generated a DCAR-Ig fusion protein, composed of the Fc portion of human IgG fused to the extracellular domain of DCAR (stalk and CRD), which bears the well-established EPN motif instead of the EPS sequence (DCAR EPN -Ig). Binding of DCAR EPN -Ig to plate-coated AcPIM2 was significantly reduced when compared with DCAR-Ig (Fig. S2A). An even stronger decrease in signaling activity was observed when stimulating reporter cells expressing the DCAR EPN mutant with AcPIMs (Fig. S2B). These results thus indicate that the EPS sequence is indeed a Ca 2ϩ -binding motif and necessary for optimal ligand binding and signaling through DCAR. Furthermore, an additional electron density corresponding to a glycerol molecule is observed adjacent the Ca 2ϩ coordination site (Fig. 3A, left panel), and the comparison with the location of trehalose in the structure of Mincle (PDB code 4KZV) showed that the two OH groups of glycerol were in a similar position to that of the 3-and 4-OH groups of trehalose ( Fig. 3B) (14). Furthermore, we also observed two H 2 O molecules in the same position in the ligand-free DCAR crystals grown in a glycerol-free condition (Fig. S3A). These results suggest that DCAR, through its atypical EPS motif, interacts with the sugar moiety of its ligands.

Ligand recognition requires the short and neutral residue Ala 136
Unlike the conserved residues of the Ca 2ϩ coordination site, the ones surrounding it are rather different between DCAR and Mincle (Fig. 3B). Among these, we first assessed the role of the Ala 136 (Glu 135 in Mincle) by mutagenesis. Because DCAR ligands bear a negatively charged phosphate (7), we surmised that, contrarily to Mincle, the non-negative and relatively short side chain of Ala 136 was required to accommodate a phosphate moiety in this area. We therefore introduced a negatively charged Glu (the corresponding residue of Mincle; DCAR A136E ) or Asp (DCAR A136D ) at this position and induced the expression of the mutant receptors in reporter cells. The presence of a negatively charged residue at this position as is the case for Mincle, abolished GFP expression in response to ligand stimulation ( Fig. 3C), implying that a negative charge in the vicinity of the primary sugar-binding site may prevent the interaction with AcPIMs. To examine the impact of the size of the side chain, we substituted the neutral Ala 136 with neutral residues of increasing side-chain length (Val Ͻ Leu/Ile). Although the mutant with the shortest residue (DCAR A136V ) retained significant reporter activity, the DCAR A136L and DCAR A136I mutants showed strongly reduced activation (Fig. 3D). These results were confirmed by the introduction of these mutations in DCAR-Ig fusion proteins, which had a similar effect on direct binding to plate-coated AcPIMs ( Fig. 3E and Fig. S2C). It thus suggests that the Ala 136 is required to avoid electrostatic and steric hindrance with the phosphate moiety of AcPIMs.

Critical role of Gln 198 for ligand binding
The residues corresponding to the Gln 198 of DCAR (Fig. 3B) in mDCIR2, hDCIR, and hDectin-2 (His 225 , Gln 226 , and Arg 198 , respectively) were shown to be part of a secondary sugar-binding site in these three closely related CLRs (13,15,16). Consistent with these results, Gln 198 was shown to be required for DCAR, because substitution with Ala and the slightly shorter Asn both abolished binding efficiency (Fig. S2D). Moreover, the strong reduction of binding induced by introducing a Glu at , with identical residues highlighted in red and similar residues enclosed in blue boxes (m, mouse; c, cow; h, human). Secondary structure elements (␣-helices and ␤-strands) of DCAR are shown above the primary structures, and each pair of cysteines involved in disulfide-bond formation are designated by a number below the sequences. The green filled circles indicate the residues participating in Ca 2ϩ coordination, and the residues involved in ligand binding are enclosed in black boxes. Sequence alignment was obtained with ESPript (33).

Crystal structure of C-type lectin receptor DCAR
this position confirmed the detrimental effect of a negative charge near the primary sugar-binding site, as seen for Ala 136 (Fig. S2D).
In hDCIR, hBDCA-2, and mDCIR2, an Asn corresponds to the Tyr 180 of DCAR and interacts with the carbonyl group of the N-acetylglucosamine (GlcNAc) at the nonreducing end of mannose in all three structures (13,15,17,18). Mutation of Tyr 180 led to a partial reduction of reporter activity (Fig. S2E), suggesting that this residue is also involved in binding, but to a lesser extent than the Gln 198 . Collectively, these results may therefore imply that DCAR binds an additional sugar of its ligands through a secondary sugar-binding site.

DCAR binds the phosphosaccharide moiety of AcPIMs
Among CLRs, DCAR has the uncommon characteristic of recognizing and binding phosphoglycolipids from mycobacteria such as AcPIM2 (Fig. 4A). We thus sought to investigate whether the sugar and phosphate moiety interacts with DCAR. We generated a soluble synthetic compound that comprises the phosphosaccharide moiety of AcPIM2 without lipid chains (inositol-monophosphate dimannose; IPM2), which can be used in aqueous binding assays (Fig. 4A, shaded in blue). We first attempted to inhibit binding of the DCAR-Ig fusion protein to plate-coated AcPIM2 by this analog (Fig. 4A). Contrary to free D-mannose, this compound induced a dose-dependent inhibition of DCAR-Ig binding to its ligand. These results suggest that DCAR has the potential to interact with at least mannose linked to inositol monophosphate.
To measure direct binding of DCAR-Ig to the analog IPM2, we turned to the sensitive biolayer interferometry (BLI) assay. Neither free mannose nor free D-myo-inositol (data not shown) could be bound by DCAR-Ig. Consistent with the competitive binding assay shown in Fig. 4A, real-time binding of DCAR-Ig, but not Ig, to the synthetic PIM analog was observed by BLI (Fig. 4B, left panels). In this experimental setting, steady-state analysis (Fig. 4B, right panels) demonstrated a dissociation constant (K d ) of 140 Ϯ 7.9 M. This result therefore indicated that within the whole ligand, DCAR has substantial affinity with the phosphosaccharide moiety.

DCAR complex structure with IPM2
We finally tried to verify this interaction by obtaining the structure of DCAR in complex with IPM2 (Fig. 5A). Co-crystals of the DCAR CRD-IPM2 complex were diffracted up to 1.84 Å Crystal structure of C-type lectin receptor DCAR resolution in the P2 1 2 1 2 1 space group, and two complexes were in the asymmetric unit. The electron density of IPM2 was clearly identified except for one mannose (Fig. 5B). The structure shows that DCAR interacts with mannose through the 3and 4-OH groups (Fig. 5A), which confirms our assumption raised from the interaction with glycerol (Fig. 3A) and that DCAR binds mannose through its EPS motif. Importantly, we did not observe apparent conformational change upon ligand binding, which is often seen in "non-CLR" immune receptors (19,20). Rather, ligand-free and IPM2-bound structures of DCAR were almost identical (Fig. 5, C-E), supporting the idea that CLRs do not utilize conformational changes as a strategy for signaling.

Contribution of a hydrophobic groove to the interaction with AcPIMs
Our analysis of DCAR binding to IPM2 by BLI showed that the interaction dissociated rapidly (Fig. 4B). This prompted us to determine whether another determinant may be involved in the interaction with the whole "acylated" ligands AcPIMs. The crystal structure revealed a hydrophobic area close to the IPM2-binding site (Fig. 6, A and B), which is in a suitable position to accommodate the acyl chains of AcPIMs. We therefore reduced the hydrophobicity of this area by substituting the Ile 133 by Gly (DCAR I133G ) or Ala (DCAR I133A ). The DCAR reporter activity was strongly reduced for both mutants (Fig.  6C), whereas these mutations did not grossly disrupt the receptor structure and function (Fig. 6D). These results thus suggest that this hydrophobicity also contributes to the recognition of the mycobacterial AcPIMs via its receptor DCAR.

Discussion
We report here the atomic details of unique ligand recognition mode of the CRD of DCAR toward phosphoglycolipids. The Ca 2ϩ -binding motifs of C-type lectins were shown to directly influence sugar specificity. Indeed, substituting the and Mincle (pink, right panel, PDB code 4KZV) bound to glycerol (green) and trehalose (yellow), respectively. Hydrogen and Ca 2ϩ coordination bonds are depicted by black dotted lines. B, superimposed structures of the residues adjacent to the putative binding site of glycerol-bound DCAR (cyan) and trehalose-bound Mincle (pink). C and D, reporter cells expressing DCAR or its mutants were stimulated with increasing amounts (30, 100, and 300 ng/well) of plate-coated Ac 2 PIM2 for 18 h, and NFAT-GFP activity was assessed by flow cytometry. E, the Fc portion of human IgG (Ig), WT or mutant DCAR-Ig fusion proteins were incubated with increasing amounts (10, 30, and 100 ng/well) of plate-coated Ac 2 PIM2, and binding was detected using anti-hIgG-horseradish peroxidase. The data are presented as the means Ϯ S.D. of triplicate assays and are representative of three independent experiments (C-E).

Crystal structure of C-type lectin receptor DCAR
EPN motif of rat MBP-A (mannose-binding protein A) by QPD imparts galactose binding ability to this C-type lectin (12). Although the EPN and EPS motifs both have previously been shown to bind mannose moieties (5, 13), here we show that having an EPN motif in lieu of EPS significantly reduces ligand recognition by DCAR. Interestingly, the amine group of Asn 170 of EPN 168 -170 makes a hydrogen bond with the 3-OH of glucose in the Mincle-trehalose complex (Fig. 3A, right panel), but the Ser 170 in EPS 168 -170 of DCAR does not have this hydrogen donor ( Fig. 5A and Fig. S3B). It may therefore be plausible that the EPS motif sets the ligand in a different position than the EPN, which could contribute to efficient ligand binding by DCAR. Thus, the C-type lectins may have evolved to recognize a broader spectrum of ligands by mutational changes in their binding motifs while retaining their overall canonical fold. Indeed, several unconventional EPN-like motifs, such as EPS and EPK, are found in C-type lectins across species (5,10,21,22). DCAR belongs to the DCIR family of CLRs, which share structural and ligand-binding similarities. Among them, the secondary binding site of hBDCA-2 and mDCIR2 interacts with GlcNAc linked to mannose (13,15,17). The Asn 184 and Arg 186 of hBDCA-2 interact with the carbonyl group of GlcNAc. Although the Arg (residue 182) is conserved in DCAR, the res-

Crystal structure of C-type lectin receptor DCAR
idue corresponding to the Asn of hBDCA-2 is Tyr 180 . Therefore, the hydrogen bond network with GlcNAc seen in hBDCA-2 would not be feasible for DCAR. mDCIR2 also binds GlcNAc through two additional residues (Asp 223 and His 225 ), but they are not conserved in DCAR. This shows that although DCAR shares structural similarities with DCIR family members, it possesses fine characteristic structural features to achieve its unique specificity.
The loop region between the ␤2 and ␤3 strands of mDCIR2 was suggested to create a narrow carbohydrate-binding site when compared with dendritic cell-specific ICAM-3 grabbing nonintegrin (DC-SIGN), which would lead to lower affinity for high mannose ligands. It was also proposed that this may be a conserved feature among DCIR family members (15). Interestingly, we have previously shown that among the AcPIM species, which vary in their number of mannose residues, DCAR activation and binding weakens as the number of mannose increases (Fig. S4A), whereas the opposite tendency was observed for DC-SIGN (Fig. S4B) (7). Because the ␤6-␤7 loop of DCAR also generates a narrow binding site when compared with that of DC-SIGN (Fig. S4C), it could explain DCAR poor binding to AcPIM6 and Ac 2 PIM6. Our data therefore corroborate this hypothesis, and the characteristic narrow carbohydrate-binding site may indeed be shared by DCIR family CLRs, which may contribute to their specificity for low mannose ligands.
Our complex analysis clearly demonstrated the interaction of mannose to DCAR. Although inositol and phosphate were not observed in the proximity of the CRD from the co-crystal structure with IPM2, it does not exclude the possible contribution of phosphate and inositol in the context of whole AcPIMs. Indeed, our BLI analysis suggests that DCAR can interact with the phosphosaccharide IPM2 but not with mannose alone. Similarly, IPM2, but not monosaccharide, was able to block the interaction between DCAR and AcPIM2 (Fig. 4A).
However, the interaction of DCAR CRD and IPM2 was transient and dissociated rapidly, implying that another additional determinant also contributes to stable interaction, which may be required for optimal signal transduction through DCAR. Mincle was shown to have a hydrophobic region close to the primary sugar-binding site that allows the accommodation of the lipid moiety of its ligands (11,23). The structure of DCAR also shows a hydrophobic surface, although in a different area than that of Mincle, extending from the primary sugar-binding site (Fig. 6, A and B), which also contributed to the AcPIM binding (Fig. 6C).
Acylated PIMs may have up to four acyl chains; there are two within the phosphatidyl moiety, and for AcPIM2, one additional chain is linked to the 6-OH of the Man2 residue (Fig. 4A). Based on the DCAR-IPM2 complex, one mannose residue is bound at the primary sugar-binding site, and its 6-OH would be in a suitable position to allow an acyl chain to interact with the Crystal structure of C-type lectin receptor DCAR hydrophobic groove (Fig. 6B). Additionally, the position of the phosphate group of IPM2 suggests that the phosphatidyl moiety may also be bound by the hydrophobic area, which could bring the phosphate and inositol within interacting distance of Ala 136 , Tyr 180 , and Gln 198 . This may explain the strong reduction of ligand binding by mutation of these residues observed in response to whole AcPIMs. Thus, these two lectins for mycobacteria may have independently acquired the structural feature conferring the ability to recognize not only sugars but also lipids, because the primary structure corresponding to the putative hydrophobic grooves were different between Mincle and DCAR.
Signaling through C-type lectin receptors requires their multimerization by the recognition of multivalent ligands, thereby inducing the phosphorylation of ITAM within their signaling subunits (24,25). Therefore, theoretically these receptors do not necessarily require ligand-induced conformational change, which is usually a critical process for other receptors, such as cytokine receptors or G protein-coupled receptors. Indeed, we did not observe any drastic conformational changes of DCAR CRD after IPM2 binding (Fig. 5, C and D). Thus, although DCAR is able to recognize one ligand molecule, AcPIMs can signal through DCAR only when it is presented on the bacteria at high valency. This may be an additional strategy to discrim-inate real pathogens from the presence of structurally similar glycolipids that should not trigger immune responses in the host. Indeed, IPM2 did not activate cells expressing DCAR (data not shown), although it bound to DCAR (Fig. 4B).
The structural analysis of DCAR revealed particular features allowing this receptor to bind phosphate-containing ligands, implying the presence of other pathogen-or host-derived phosphoglycolipids recognized by this CLR. Because no homolog of DCAR was found in the human Dectin-2 cluster, the structure of mouse DCAR may contribute to a rational structural approach toward the identification of a functional homolog or unknown receptor(s) for AcPIMs in human.

Recombinant protein expression and purification
The CRD of mouse DCAR (residues C78-L209) was inserted in the E. coli expression plasmid pGMT7 (primers are shown in Table S1), which then was transformed into the E. coli strain

Crystal structure of C-type lectin receptor DCAR
Rosetta TM (DE3) (Novagen). Because of a cloning artifact, three additional amino acids were expressed at the N-terminal end upstream of the C78. Transformants were grown in selective LB medium, protein expression was induced with 0.5 mM isopropyl ␤-D-thiogalactoside (Nacalai Tesque Inc.), and the cells were cultured for 3 h. The cells were then harvested at 4000 rpm for 20 min, resuspended in lysis buffer (10 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 150 mM NaCl, 10% glycerol) and disrupted by sonication. After a 30-min DNase I (Roche, 100 g/ml) treatment, inclusion bodies were purified by centrifugation. Inclusion bodies were solubilized in a buffer containing 6 M guanidine, 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM EDTA. DTT (10 mM) was then added to solubilized proteins, which were then slowly incorporated into a refolding buffer containing 100 mM Tris-HCl, pH 8.0, 2 mM EDTA, 0.4 M L-arginine, 6.5 mM cysteamine, 3.6 mM cystamine and slowly stirred at 4°C for at least 3 h. The proteins were then dialyzed twice against 10 mM Tris-HCl, pH 8.0, and purified by anion exchange (Poros 50 HQ column, Applied Biosystems) and size-exclusion (Superdex TM 200 10/300 column, GE healthcare) chromatography. The fractions corresponding to the DCAR CRD were pooled and enriched using an Amicon Ultra centrifugal filter (molecular weight cutoff: 3000, Merck Millipore) for crystallization assays.

Crystallization and structure determination
All crystals were generated by crystallizing purified recombinant DCAR (10 mg/ml for ligand-free and DCAR-IPM2 complex and 13 mg/ml for DCAR-glycerol complex) using the sitting-drop vapor-diffusion method at 20°C. DCAR-glycerol complex crystals were obtained by growing crystals for 4 days under a reservoir solution consisting of 0.04 M citric acid, 0.06 M Bis-Tris propane, pH 6.4, 20% (w/v) PEG 3350, 10 mM CaCl 2 . For data collection, the crystals were transferred to the cryo buffer (0.04 M citric acid, 0.06 M Bis-Tris propane, pH 6.4, 22% (w/v) PEG 3350, 10 mM CaCl 2 , 15% glycerol). Ligand-free and DCAR-IPM2 complex crystals were grown for 7 days under a reservoir solution consisting of 0.04 M citric acid, 0.06 M Bis-Tris propane, pH 6.4, 35% (w/v) PEG 3350, 10 mM CaCl 2 , with or without 120 mM IPM2. For data collection, the crystals were frozen directly from the drop in liquid nitrogen. Thereafter, all crystals were flash-frozen using a cryo system (Rigaku). X-ray diffraction data sets were collected on Beamline BL45XU at SPring-8 (Hyogo, Japan). The diffraction data were processed using the program package XDS (26). The crystals were found to belong to space group P2 1 2 1 2 1 . The data collection statistics are compiled in Table 1. The crystal structure of the DCAR-glycerol complex was determined by molecular replacement using Molrep (27) with the template structure (PDB code 3VYJ), and the ligand-free and IPM2-complex structures were obtained by molecular replacement using the glycerol-complex structure. The structures were manually modified using the Coot program (28). Several iterative rounds of model building were performed in Coot and refined using Refmac5 (29). Refinement statistics are summarized in Table 1. The quality of the structures was assessed using PROCHECK (30).

Synthesis of IPM2
The IPM2 analog having D-myo-inositol 1-phosphate was chemically synthesized from D-mannose and commercially available D-glucose derivative. First, D-myo-inositol acceptor was prepared from methyl ␣-D-glucopyranoside via 10 reaction steps according to the literature (31). Moffat oxidation of the primary alcohol of 2,3,4-tri-O-benzyl-protected glucoside, which was derived from methyl ␣-D-glucopyranoside via three steps, followed by treatment with acetic anhydride provided the Z-olefin product in good yields. Subsequent treatment of the olefin with palladium chloride afforded the 3-OH penta-oxygenated cyclohexanone derivative. The reduction of the resulting cyclic ketone with Me 4 NBH(OAc) 3 provided the diol derivative of D-myo-inositol, which was then converted into the inositol diol acceptor with the allyl group at the C-1 position via three steps. On the other hand, 2-O-acetyl-3,4,6-tri-O-benzyl-␣-D-mannopyranosyl trichloroacetimidate donor to be coupled with the inositol acceptor was readily prepared from D-mannose via seven steps. The glycosylation of the inositol acceptor with the mannose donor was performed in the presence of trimethyl trifluoromethanesulfonate in diethyl ether at 0°C, affording the dimannoside in good yield and ␣-selectivity. Removal of the allyl group was then conducted by treatment of the dimannoside with Ir complex under hydrogen atmosphere in THF followed by hydrolysis with aqueous HCl, giving the alcohol in 93% yield. The phosphoester linkage was introduced to the C-1 position of the inositol residue by coupling with dibenzyl N,N-diisopropylphosphoramidite and 1H-tetrazole in CH 2 Cl 2 and subsequent m-CPBA oxidation, affording the fully protected IPM2 analog in 84% yield. The final steps involved deacetylation and hydrogenolytic debenzylation to furnish the target IPM2 analog in good yields. The structure of IPM2 was verified by NMR and MS (Fig. S5-S8).

Binding assay
Mouse DCAR-Ig fusion proteins (WT and mutants) were generated as previously described (32). In brief, the extracellular domain of DCAR (residues 42-209) was fused to the N-terminal end of the hIgG1 Fc region. Fusion proteins were then incubated with increasing amounts of plate-coated AcPIMs (10, 30, and 100 ng/well), and bound proteins were detected using horseradish peroxidase-labeled anti-human IgG (Jackson ImmunoResearch). After the addition of the colorimetric substrate, peroxidase activity was measured using a spectrophotometer. The mutant Ig fusion proteins were generated by recombinant PCR (primers are shown in Table S1).

Reporter assay
2B4-NFAT-GFP reporter cells were prepared as previously described (32). Briefly, we induced stable expression of FcR␥ and DCAR or its mutants (generated by recombinant PCR; primers are shown in Table S1), and the cells were stimulated with increasing amounts of plate-coated AcPIMs (30, 100 and 300 ng/well) or anti-mouse DCAR rat mAb (MAB2757, clone 349214, R&D Systems; 0.3, 1, and 3 g/ml). GFP expression was then assessed by flow cytometry.