DCIR and its ligand asialo-biantennary N-glycan regulate DC function and osteoclastogenesis

Dendritic cell (DC) immunoreceptor (DCIR, or DCIR1; gene symbol, Clec4a2) is a member of the C-type lectin receptors. DCIR contains a carbohydrate recognition domain (CRD) in the extracellular part, which can recognize mono- and oligo-saccharide structures in a Ca²⁺-dependent manner. The DCIR CRD has an EPS (glutamic acid–proline–serine) motif similar to the authentic carbohydrate recognition EPN (glutamic acid–proline–asparagine) motif that binds mannose, glucose, GlcNAc, and fucose. Several molecules, including asialo-N-glycans, were reported to bind DCIR (Bloem et al., 2014; Bloem et al., 2013; Hsu et al., 2009; Lee et al., 2011; Massoud et al., 2014; Nasu et al., 2020). However, the effects of these glycans on DCIR functions have yet to be examined.

DCIR has a canonical immunoreceptor tyrosine-based inhibitory motif (ITIM) in the cytoplasmic portion (Graham and Brown, 2009), which transduces inhibitory signals by activating phosphatases SHP-1 and SHP-2 upon stimulation (Richard et al., 2006). Previously, we showed that aged Dcir^−/− mice spontaneously develop autoimmune enthesitis and sialadenitis, because antigen presentation ability of the immune system is enhanced due to excess differentiation and proliferation of DCs in Dcir^−/− mice (Fujikado et al., 2008). Furthermore, Dcir^−/− mice are highly susceptible to collagen-induced arthritis (CIA; Fujikado et al., 2008), a model for rheumatoid arthritis (Brand et al., 2007), and experimental autoimmune encephalomyelitis (EAE; Seno et al., 2015), a model for multiple sclerosis (Stromnes and Goverman, 2006), respectively, indicating an important regulatory role of DCIR in the immune system.

We also found that ankylotic changes of the ankle joints develop in Dcir^−/− mice with age due to fibrocartilage proliferation and heterotopic ossification in joints, associated with an increase in the femoral bone mass (Fujikado et al., 2008; Maruhashi et al., 2015). This is because T helper type 1 cell differentiation is enhanced in Dcir^−/− mice, and overproduced IFN-γ from T helper type 1 cells promotes chondrogenesis and bone formation (Maruhashi et al., 2015). Thus, DCIR plays important regulatory roles in both the immune system and the bone metabolic system. However, the regulatory mechanisms of DCIR activity remain obscure.

In this study, we showed that an endogenous asialo-biantennary N-glycan (NA2) can bind DCIR. By analyzing the effects of NA2 on osteoclastogenesis, we identified NA2 as a functional ligand for mouse and human DCIR. Furthermore, we showed that exposure of NA2 on the cell surface by removing the terminal sialic acid of N-glycans with sialidases suppresses osteoclastogenesis and DC maturation by acting on DCIR on the cell surface. Sialidase treatment also suppressed the development of CIA and EAE. These findings suggest that the NA2-DCIR axis is critically important for the homeostasis of the immune and bone system.

To identify the functional ligands of DCIR, we generated DCIR-Fc, in which the extracellular domain (EC) of DCIR was fused with the Fc region of human IgG2, and examined the binding of DCIR-Fc to different types of cells. Although T and B lymphocytes did not bind to DCIR-Fc (data not shown), a portion of macrophages was positive for DCIR-Fc staining (Fig. S1, a and b). Osteoclasts, which are differentiated from the cells in the monocyte-macrophage lineage, were also partially positive for DCIR-Fc staining (Fig. S1, c and d). The binding was reduced with DCIRE197A/S199A-Fc, a mutant DCIR-Fc with two point mutations in EPS (Zelensky and Gready, 2005; Fig. S1, a and b). Moreover, since less binding was observed after treatment of macrophages with EDTA or trypsin, it was determined that DCIR interacts with glycoproteins on macrophages and osteoclasts in a CRD- and Ca²⁺-dependent manner (Fig. S1, a and b). Since osteoclast differentiation is tightly regulated by osteoblasts (Lee et al., 2010), we investigated whether DCIR ligands were presented on osteoblasts. Osteoblasts were positive for DCIR-Fc staining, and a reduction in binding occurred in the presence of EDTA (Fig. S1 e). As is the case with macrophages and osteoclasts, the binding of DCIR-Fc to osteoblasts was dependent on the EPS motif (Fig. S1 f). To further characterize the DCIR ligands, osteoblasts were treated before DCIR-Fc staining with tunicamycin, which blocks N-linked glycosylation, and benzyl-α-GalNAc, which inhibits O-linked glycosylation. The binding of DCIR-Fc was abolished after treatment with tunicamycin, but not benzyl-α-GalNAc (Fig. S1 g), indicating that DCIR ligands are likely to be N-linked glycans on glycoproteins.

Since N-linked glycans were considered to be a potential ligand for DCIR, a glycan microarray with the evanescent-field fluorescence-assisted detection system (Tateno et al., 2008) was used to screen the saccharide ligands using the DCIR-Fc protein as the probe. By exploiting this system, it was possible to detect weak interactions with high sensitivity. The glycan microarray revealed that DCIR-Fc bound to asialo-transferrin and heparin in a Ca²⁺-dependent manner, but not to glycoproteins expressing high mannose-type glycans, highly branched complex-type N-linked glycans, fucosylated glycoconjugates, or O-linked glycans that were spotted on the array (Fig. S2). DCIR-Fc bound to asialo-transferrin and heparin in a dose-dependent manner, and EDTA completely blocked the binding (Fig. 1 a). No binding of mutant DCIR-Fc (E197A-Fc, S199A-Fc, and E197A/S199A-Fc) to asialo-transferrin was observed, whereas the mutant DCIR-Fcs remained to bind heparin (Fig. 1 b). Because the binding of DCIR-Fc depended on CRD, and it was considered likely that N-linked glycans were the ligands for DCIR (Fig. S1), we focused on identifying the DCIR ligand on asialo-transferrin. Since the biantennary structure of N-glycans including NA2 is predominant on transferrin (Regoeczi et al., 1979), we tested the binding of DCIR-Fc to NA2. As shown in Fig. 1, c and d, DCIR-Fc bound strongly to NA2 in a Ca²⁺- and CRD-dependent manner. As human and mouse DCIR have 65% homology of the amino acid sequences over the whole sequence with the EPS motif in the EC, we examined the binding of human DCIR to NA2. As shown in Fig. 1 e, human DCIR-Fc also bound to NA2 in a Ca²⁺-dependent manner.

It has been shown that DCIR can bind to the glycans expressed by Chinese hamster ovary (CHO) cells (Bloem et al., 2013). Therefore, we examined DCIR-Fc binding to glycan transporter–deficient CHO cells. The binding of DCIR-Fc was strongly enhanced in Lec2 cells that expressed sialic acid–deficient complex N-glycans due to the lack of cytidine 5′-monophosphate–sialic acid transporters compared with the parental CHO cells (Fig. 1 f). However, no binding was detected on Lec8 cells, which expressed sialic acid/galactose–deficient complex N-glycans due to the lack of uridine 5′-diphophate–galactose transporters, or on the Lec1 cells, which lack complex and hybrid-type N-glycans due to defective N-acetylglucosaminyltransferase I.

To obtain further structural insight into the importance of the biantennary structure and desialylated forms, we transfected 293T cells, which expressed a low level of the DCIR ligand, with Mgat2 (β-1, 2-N-acetylglucosaminyltransferase II) to transfer GlcNAc to the β1-2 of Manα1-6 (GlcNAcβ1-2Manα1-3) Manβ1-4GlcNAcβ1-4GlcNAcβ1-Asn in order to achieve an overexpression of NA2. Although nontransfected 293T cells were not bound to DCIR-Fc, Mgat2-293T cells were clearly positive for DCIR-Fc (Fig. 1 g). These results confirm that mouse and human DCIR interact with the desialylated biantennary structures of N-glycans.

Because DCIR is highly expressed in myeloid-lineage cells, such as DCs and macrophages, and Dcir^−/− mice develop bone abnormalities, we next examined the expression of DCIR in osteoclasts, one of the myeloid lineage cells. Osteoclasts originate from hematopoietic stem cells in bone marrow (BM), and both the macrophage CSF (M-CSF) and receptor activator of NF-κB ligand (RANKL) are critically important for the differentiation of osteoclasts (Negishi-Koga et al., 2011; Okamoto and Takayanagi, 2019; Pederson et al., 2008). In addition, an immunoreceptor tyrosine-based activation motif–mediated signal is a requirement for terminal osteoclast formation, and is counterbalanced by phosphatases (Kaifu et al., 2003; Koga et al., 2004; Takeshita et al., 2002; Umeda et al., 1999). We found that DCIR was expressed in M-CSF– and RANKL-induced osteoclasts (Fig. 2 a) and M-CSF–induced BM macrophages (BMMs; Fig. S2, b and c). Its expression progressively increased during osteoclast maturation. DCIR was also detected in CD11b-positive OsteoMacs (Chang et al., 2008), but not in osteoblasts (Fig. S2 d).

We examined the role of DCIR in osteoclastogenesis in an in vitro culture system. When osteoclasts were induced to differentiate from BMMs with M-CSF and RANKL, Dcir^−/− BMMs formed tartrate-resistant acid phosphatase (TRAP)–positive osteoclasts with >10 nuclei, making them more efficient than WT BMMs (Fig. 2 b). The expression of osteoclastic genes, such as Nfatc1 (a master regulator of osteoclast formation), Acp5 (an enzymatic marker of osteoclasts), Src (a tyrosine kinase essential for bone resorption), Calcr (a receptor for calcitonin), and Ctsk (a protease responsible for degrading type I collagen), was significantly increased in Dcir^−/− osteoclasts (Fig. 2 c). Osteoclast differentiation was significantly suppressed by the retroviral expression of DCIR, but not by an ITIM mutant (DCIRY7F) that lacked the ability to deliver inhibitory signaling, indicating that ITIM-mediated signaling is required for the DCIR-induced suppression of osteoclastogenesis (Fig. 2 d). Furthermore, multinucleated osteoclast formation was also enhanced when nonadherent Dcir^−/− BM cells (BMCs) were cocultured with WT osteoblasts, but not vice versa, indicating that DCIR is a unique inhibitory receptor that down-regulates osteoclastogenesis (Fig. 2 e).

We investigated the effect of NA2, putative DCIR ligands, on osteoclastogenesis. When WT BMMs were induced to differentiate with M-CSF and RANKL in the presence of DCIR-Fc, which blocks the interaction between DCIR and its ligand, a significant increase in the number of TRAP⁺ multinucleated cells was observed (Fig. 3 a). However, a mutant form of DCIR-Fc, DCIRE197A/S199A-Fc, had no effect at all, indicating that the effect of DCIR-Fc on osteoclastogenesis depends on the CRD. DCIR-Fc had no effect on osteoclast formation from Dcir^−/− BMMs (Fig. 3 a). Consistent with this observation, multinucleated TRAP⁺ osteoclast formation was strongly inhibited by treatment with NA2 in the WT BMM culture, but not in the Dcir^−/− BMM culture (Fig. 3 b). In addition, we examined the effect of NA2 on the osteoclastogenesis of human cells. Osteoclast formation was significantly suppressed by the treatment with NA2 of human CD14⁺ cells from peripheral blood (Fig. 3 c), indicating that NA2 is also the functional ligand for human DCIR to regulate osteoclastogenesis.

Because the binding of DCIR-Fc was increased in Lec2 cells, which lack terminal sialic acid in complex N-glycans (Fig. 1 e), we next examined the effect of neuraminidase treatment, which results in more desialylated glycans on the cell surface, on the binding of DCIR-Fc. Interestingly, neuraminidase treatment of BMMs enhanced the binding of DCIR-Fc, but not DCIRE197A/S199A-Fc (Fig. 3, d and e). Furthermore, RANKL-induced osteoclast formation was drastically inhibited by the treatment with neuraminidase in WT BMMs, but no such effect was noted in Dcir^−/− BMMs (Fig. 3 f). Neuraminidase treatment also suppressed multinucleated osteoclast formation in WT cells, but not in Dcir^−/− cells in a nonadherent BMC and osteoblast co-culture system (Fig. 3 g). β-Galactosidase treatment did not affect osteoclast formation in WT cells and Dcir^−/− cells, indicating that the terminal galactose in complex N-glycans is involved in the recognition by DCIR.

We examined the effect of neuraminidase treatment on modifying bone metabolism in vivo. First, we evaluated DCIR-Fc binding to BMCs after in vivo treatment with neuraminidase. 1 d after neuraminidase administration, DCIR-Fc–positive cells in BMCs were higher than those injected with PBS (Fig. S3, a and b), showing that the neuraminidase treatment enhanced exposure of NA2 on the cell surface by removing the terminal sialic acids of N-glycans.

To examine the osteoclast activity after the intravenous injection of neuraminidase, we measured the calcium ion concentration and TRAP activity in sera. While the concentration of calcium ions and the phosphatase activity were significantly reduced in WT, this was not observed in Dcir^−/− mice after neuraminidase treatment (Fig. 3, h and i). These observations are consistent with the notion that excess NA2 exposure suppressed osteoclast activity in a DCIR-dependent manner. Thus, we conclude that NA2 is a functional and specific DCIR ligand that negatively regulates osteoclastogenesis.

Since DCIR negatively regulates the development of CIA (Fujikado et al., 2008), we analyzed the effect of administering neuraminidase on the development of this disease. We used DBA/1J mice, because these mice are more susceptible to CIA than C57BL/6J mice. When CIA was induced after treatment with neuraminidase, the arthritic score was significantly lower than that of PBS-treated mice (Fig. 4 a). The cumulative arthritic score and mean maximal score were also suppressed, while the incidence, mean maximal score, and disease onset time were not significantly affected in neuraminidase-injected mice (Fig. 4 b). We also assessed the histological parameters. The reduction in the inflammatory score, pannus formation, cartilage erosion, and bone destruction was much more pronounced in neuraminidase-treated mice than in PBS-treated mice (Fig. 4, c and d). A significant decrease in the serum concentration of IFN-γ and IL-17A, but not of IL-6, was noted at the end of clinical monitoring in neuraminidase-treated mice (Fig. 4 e). To examine the effect of neuraminidase on bone resorption during CIA, we measured the serum concentration of type I collagen C-terminal telopeptide (CTX-I), which is produced by cathepsin K from type I collagen. The concentration of CTX-I in the serum of neuraminidase-treated DBA/1J mice was significantly lower than in PBS-treated mice (Fig. 4 f). Because osteoclasts mediate bone destruction in CIA, TRAP⁺ cells were quantified in joint sections. Neuraminidase treatment significantly decreased osteoclast numbers in joints (Fig. 4, g and h).

We next analyzed the effects of neuraminidase treatment on Dcir^−/− mice to examine dependency on DCIR. When neuraminidase was injected intravenously into C57BL/6J mice, the arthritic scores, but not the incidence of arthritis, of neuraminidase-treated WT mice were significantly lower than those of PBS-treated WT mice, whereas the injection had no effect on the arthritic scores of Dcir^−/− mice (Fig. 5 a). The cumulative arthritic score and mean maximal score, but not the disease onset time, were also significantly suppressed, compared with those of PBS-treated mice (Fig. 5 b). Histological assessments showed that both leukocyte infiltration and pannus formation were ameliorated in neuraminidase-treated WT mice compared with PBS-treated WT mice, whereas both parameters were high in Dcir^−/− mice irrespective of neuraminidase treatment (Fig. 5, c and d). The scores for cartilage and bone damage in neuraminidase-treated WT, but not Dcir^−/−, mice showed a similar tendency, although the differences were not statistically significant (cartilage damage, P = 0.86; bone damage, P = 0.45; Fig. 5, c and d). Also, no significant change was observed in the anti-collagen antibody titers in these mice (Fig. S3 c). Furthermore, serum concentration of CTX-I in neuraminidase-treated WT mice was lower than in PBS-treated WT mice, but the injection had no effects on the CTX-I concentration of Dcir^−/− mice (Fig. 5 e). Consistent with this, TRAP⁺ cells were more significantly decreased in neuraminidase-treated WT mice than those of PBS-treated WT mice (Fig. 5, f and g). These results suggest that neuraminidase treatment causes excess exposure of NA2 and attenuates the development of CIA through the activation of DCIR.

In an earlier study by our group, we reported that the deficiency of DCIR exacerbates the development of EAE, indicating that DCIR is involved in the suppression of autoimmunity (Seno et al., 2015). Then, we examined the effects of neuraminidase treatment during the development of EAE. Neuraminidase injection in C57BL/6J mice significantly ameliorated the development of EAE compared with PBS-treated WT mice, and this treatment had no effect on Dcir^−/− mice (Fig. 6 a). Cumulative clinical score and maximum score were significantly reduced in neuraminidase-treated WT mice, but disease onset time was comparable between neuraminidase- and PBS-treated mice (Fig. 6 b). Histological analysis also showed that leukocyte infiltration was reduced in neuraminidase-treated WT mice, while leukocyte infiltration in Dcir^−/− mice was comparable in both neuraminidase- and PBS-treated mice (Fig. 6 c). Histological score in neuraminidase-treated WT mice was lower than in PBS-treated WT mice (Fig. 6 d). Moreover, Kluver–Barrera (KB) staining showed that the state of demyelination, which is weaker pale blue area containing blue violet dots in the white matter, was significantly ameliorated in neuraminidase-treated WT mice (Fig. 6, e and f). However, neuraminidase treatment didn’t mitigate demyelination in Dcir^−/− mice, indicating that interaction of DCIR with NA2 alleviates immune responses to self-antigens.

Since immune responses were suppressed by excess DCIR signaling with neuraminidase treatment, we next examined the effect of neuraminidase treatment on DC-mediated T cell response. To determine whether DCs possess DCIR ligands, we differentiated DCs from BMCs with GM-CSF (GM-DCs) and evaluated the DCIR-Fc binding to GM-DCs, which were treated by neuraminidase before the biding assay. DCIR-Fc binding to GM-DCs was significantly increased after neuraminidase treatment, indicating that neuraminidase treatment enhanced ligand exposure on GM-DCs (Fig. 7 a). However, GM-DC cell number did not change between neuraminidase-treated and nontreated DCs during differentiation, suggesting that neuraminidase treatment did not affect GM-DC growth or survival (Fig. 7 b).

We analyzed the role of DCIR–NA2 interaction in DC activation during DC-mediated T cell activation. For this, we exploited 2D2 TCR transgenic mice (2D2 mice) that express T cell receptors specific for a myelin oligodendrocyte glycoprotein amino acid 35–55 peptide (pMOG). Splenic DCs purified from WT or Dcir^−/− mice were co-cultured with Thy1.2-positive T cells from 2D2 mice in the presence of pMOG, and gene expression was analyzed by RNA sequencing (RNA-seq). We found that the expression of 460 genes in the Dcir^−/− DC-2D2 T cell co-culture was significantly higher than that in the WT DC-2D2 T cell co-culture (Fig. 7 c). To understand the functional features of these genes, a gene ontology (GO) analysis was performed. Interestingly, the gene sets associated with antigen presentation (H2-Aa, Cd80, Cd86, and Cd40) were significantly higher in the Dcir^−/− DC co-culture, suggesting that DCIR-mediated inhibitory signal regulates antigen-presenting activity of DCs (Fig. 7 c).

We also found that the expression of cytokines such as Il1a, Il1b, Il21, and Il1f9 and chemokines such as Cxcl3, Cxcl2, and Ccl2 in 2D2 T cell–DC co-culture was highly augmented in Dcir^−/− DCs (Fig. 7 d). In addition, the expression of other genes was also augmented, such as those associated with pattern recognition receptor signaling pathway (Lyn, Tlr7, Ly96, and Hck), protein phosphorylation (Cd40, Tnfsf11a, and CD36), cell migration (Slamf8, Pla2g7, and Myo1f), and transcription factors (Aim4, Tnfaip3, and Acod1).

Because the expression of genes involved in antigen presentation was augmented, we next analyzed the expression of these molecules on CD11c⁺ cells in the DC-2D2 T cell co-culture by FACS. The expression of CD86 was reduced in neuraminidase-treated DCs cocultured with 2D2 T cells for 3 d in the presence of pMOG (Fig. 7 e). In this culture, the proportion of WT CD11c⁺ cells and the frequency of I-A/I-E⁺, CD86⁺, CD80⁺, and Ki67⁺ cells in CD11c⁺ cells were also significantly decreased after neuraminidase treatment compared with those in nontreated cells (Fig. 7 f). It should be noted, however, that the expression of these molecules was not affected by neuraminidase treatment in Dcir^−/− DCs.

Since a set of genes related to antigen presentation was augmented in Dcir^−/− DC-2D2 T cell co-culture, we examined antigen-presenting ability of WT and Dcir^−/− DCs with or without neuraminidase treatment. Splenic DCs purified from WT or Dcir^−/− mice were co-cultured with 2D2 T cells in the presence of pMOG, and the effects of neuraminidase treatment on T cell proliferation and cytokine production were examined. The extent of T cell proliferation in the presence of neuraminidase was lower in the co-culture with WT DCs than that in the absence of neuraminidase (Fig. 7 g). In contrast, neuraminidase treatment did not affect T cell proliferation with Dcir^−/− DCs. Moreover, neuraminidase treatment resulted in a decrease in cytokine expression, such as GM-CSF, IFN-γ, and TNF-α, but not IL-17, and Ki67 expression in T cells in the co-culture with WT DCs (Fig. 7 h; and Fig. S4). However, neuraminidase treatment did not affect a cytokine-positive T cell population in Dcir^−/− DC co-culture. We further examined the effect of NA2 on antigen-presenting ability of WT and Dcir^−/− DCs. Consistent with the observation of neuraminidase treatment in WT DC-WT T cell co-culture, NA2 decreased the expression of cytokines, such as GM-CSF, IFN-γ, and TNF-α, but not IL-17, and Ki67 in T cells (Fig. 7 i). NA2 treatment in Dcir^−/− DC-WT T cell co-culture had no effect on cytokine-positive T cell population (Fig. 7 i).

Finally, we activated 2D2 T cells with anti-CD3 and anti-CD28 antibodies in the presence of neuraminidase to examine the direct effects of neuraminidase on T cells. As shown in Fig. S5, a and b, cytokine-positive T cells and cytokine concentrations in the culture supernatants were comparable irrespective of neuraminidase treatment. These results indicate that T cell activation is suppressed indirectly through interaction with DCs, suggesting that NA2-induced DCIR stimulation suppresses the antigen-presenting ability of DCs.

In this report, we have identified NA2 as an endogenous functional ligand for DCIR. DCIR expressed on osteoclasts and DCs binds NA2 on macrophages, osteoclasts, osteoblasts, and DCs in a CRD- and Ca²⁺-dependent manner. The treatment of WT BMCs, but not of Dcir^−/− BMCs, with NA2 suppressed the differentiation of osteoclasts, indicating that NA2 is a functional ligand for DCIR. Furthermore, BMCs from Dcir^−/− mice differentiated more efficiently into osteoclasts than WT BMCs upon stimulation with M-CSF and RANKL, suggesting that NA2 on BMMs regulates their differentiation into osteoclasts. Interestingly, neuraminidase treatment, which forces an increase in NA2 exposure on the cell surface by removing the terminal sialic acids of N-glycans, suppressed the RANKL-induced osteoclastogenesis of BMMs. Since neuraminidase treatment had no observable effect in Dcir^−/− osteoclasts, it is reasonable to assume that the suppression was mediated by DCIR. These data suggest that a large proportion of NA2 is masked by sialic acids under physiological conditions and becomes available after neuraminidase treatment. Human DCIR also binds NA2 in a Ca²⁺-dependent manner, and NA2 suppressed human CD14⁺ cell differentiation into osteoclasts. We do not know yet whether this interaction of NA2 with DCIR occurs in cis, trans, or both manners on BMMs.

Consistent with our observation, it was reported that human DCIR binds biantennary N-glycans (Hsu et al., 2009), and GlcNAcβ1-2Man of biantennary N-glycans are the main epitope for the binding (Nagae et al., 2016). Although the binding of DCIR to fucose, mannose, sulfated LacNAc, Lac, Lewis^b, and Man₃ was reported using DCIR-Fc (Bloem et al., 2013; Hsu et al., 2009; Lee et al., 2011), we could not reproduce the results, probably because the binding affinity was too low (Fig. S2). We were not able to test for the binding to Man₃ because the glycanarray we used did not include Man₃. However, none of these molecules was shown to induce the ITIM-mediated biological function of DCIR.

In support of our conclusion that NA2 is a ligand for DCIR, the binding of DCIR-Fc to 293T cells was enhanced by the expression of Mgat2, which catalyzes biantennary N-glycan synthesis, and is inhibited by tunicamycin, a N-glycan synthesis inhibitor, but not by the O-glycan synthesis inhibitor benzyl-α-GalNAc. Furthermore, the cytidine 5′-monophosphate–sialic acid transporter-deficient Lec2 cells showed higher levels of DCIR-Fc binding, and uridine 5′-diphophate–galactose-transporter–deficient Lec8 cells did not bind DCIR-Fc.

We showed that the development of CIA and EAE was suppressed by treatment with neuraminidase in WT mice, but not in Dcir^−/− mice, suggesting that excess exposure of NA2 suppresses the development of these diseases in a DCIR-dependent manner. These results are consistent with our previous observations that the development of CIA and EAE is exacerbated in Dcir^−/− mice (Fujikado et al., 2008; Seno et al., 2015), although in the present experiments, exacerbation of CIA and EAE in Dcir^−/− mice was not observed. This is because we used more severe disease induction conditions than those used in the previous experiments in order to see suppressive effects of neuraminidase more clearly, and the disease severity scores were saturated even in WT mice.

Then we investigated the mechanism of how DCIR signaling regulates development of CIA and EAE. We found by RNA-seq analysis that the expression of several groups of genes was augmented in Dcir^−/− DCs co-cultured with pMOG-specific 2D2 T cells in the presence of pMOG. One of these groups was categorized as the genes involved in antigen presentation, including the H2-Aa, Cd80, Cd86, and Cd40 genes. Among them, the major histocompatibility complex H2 is important for antigen presentation, and CD80 and CD86 are important coreceptors for T cell activation. CD40 signaling in DCs induces cytokine production, such as IL-1β and TNF, and promotes antigen presentation through activation of the TRAF–NF-κB and/or MAPK pathways (Ma and Clark, 2009). Thus, these results show that NA2-induced DCIR signaling suppresses antigen presentation by DCs.

The RNA-seq analysis also revealed that the expression of gene sets relating to pattern recognition receptor signaling pathways (Lyn, Tlr7, Ly96, and Hck), protein phosphorylation (Cd40, Tnfsf11a, and CD36), cell migration (Slamf8, Pla2g7, and Myo1f), and transcription (Tnfaip3, A20, Acod1, and Id1) were also activated in Dcir^−/− mice. Thus, these observations indicate that not only the genes involved in antigen presentation but also a broad range of gene expression are augmented in Dcir^−/−DC and T cell co-culture, probably reflecting an excessively activated state of DCs.

In support of the RNA-seq results, we found by FACS analysis that the expression of molecules involved in antigen presentation of DCs was reduced in neuraminidase-treated DCs cocultured with 2D2 T cells. 2D2 T cell proliferation in the presence of pMOG and WT DCs, but not Dcir^−/− DCs, was suppressed by treatment with neuraminidase. Cytokine production from 2D2 T cells upon incubation with pMOG and DCs was also suppressed by neuraminidase treatment. These results indicate that antigen-presenting activity of DCs is regulated by the DCIR-NA2 interaction. Therefore, we conclude that the development of CIA and EAE is suppressed by the treatment with neuraminidase, because antigen presentation by DCs is reduced. Although the content of NA2 is likely to be regulated by the balance between sialyltransferases and neuraminidases, the detailed regulatory mechanisms have yet to be elucidated. While the inhibitory effect of neuraminidase treatment on in vitro RANKL-induced osteoclast formation was reported by Takahata et al. (2007), the molecular mechanism was not elucidated.

DCIR-dependent amelioration of CIA and EAE by the treatment with neuraminidase was rather unexpected for us, because sialic acids decorating the terminus of carbohydrate chains on glycoproteins and glycolipids are among the most important signaling molecules and are recognized by many receptors. For example, Siglec family members are known to be the receptors for sialic acids (Crocker et al., 2007). The expression of Siglec-15 is up-regulated during RANKL-induced osteoclastogenesis and facilitates osteoclast differentiation (Kameda et al., 2013). Deficient mice for Siglec-2 show reduced mature B cells and thymus-independent B cell responses (Nitschke, 2005). Siglec-F–deficient mice show enhanced eosinophilia in allergic airway inflammation (Zhang et al., 2007), and B1a cells are increased in Siglec-G–deficient mice (Hoffmann et al., 2007). The selectin family members are another category of receptors whose activities are regulated by the sialylation of the ligands. Leukocyte rolling, neutrophil recruitment into the inflamed peritoneal cavity, and lymphocyte homing to secondary lymphoid organs are impaired in mice lacking two sialyltransferases (St3gal4/St3gal6), which are important for the synthesis of E- and P-selectin ligands (Yang et al., 2012). The activity of other lectins that recognize terminal Gal, such as galectins (Thiemann and Baum, 2016), may also be affected by the neuraminidase treatment. Nonetheless, we observed only marginal effects, if any, on other lectins, which were considered likely to be affected by the treatment with neuraminidase. These observations suggest that DCIR is of primary importance among the C-type lectins and other lectins for the regulation of the immune system and bone metabolism under our experimental conditions.

To conclude, we have shown that NA2 is a functional ligand for DCIR, and its activity is regulated by the modification of the terminal sialic acids to regulate DC maturation and osteoclastogenesis. However, it still remains to be elucidated where and how the terminal sialic acid content of N-glycans is regulated. In addition, the roles of the DCIR-NA2 axis in humans also remain to be elucidated, although we found that NA2 regulates human osteoclast differentiation in vitro. The findings of this study have potential for use in the development of new therapeutics for the treatment of immune and bone metabolic diseases.

Dcir^−/− mice were generated as described (Fujikado et al., 2008), and these mice were backcrossed to C57BL/6J for 12 generations. 2D2 TCR transgenic (2D2 Tg) mice were purchased from The Jackson Laboratory. All the mice used in these experiments were 8–12 wk old, and these age- and sex-matched control C57BL/6J and DBA/1J mice were purchased from Japan SLC, Inc. All the mice were housed in specific pathogen–free animal rooms of the Research Institute for Biomedical Sciences, Tokyo University of Science, and the Tohoku Medical and Pharmaceutical University. Animal experiments were performed following the guidelines of each university and were approved by the Animal Experimental Committee of the Tokyo University of Science and the Animal Experiments Committee of Tohoku Medical and Pharmaceutical University.

HEK293T, CHO cells, and CHO mutants with defective glycosyl-transferases (Lec1, Lec2, and Lec8, kindly provided by J. Hirabayashi; Stanley et al., 1975) were cultured in DMEM (Gibco, Life Technologies) containing penicillin (100 U/ml), streptomycin (100 µg/ml), and 10% heat-immobilized FBS.

To differentiate BM-derived macrophages, BMCs were isolated by flushing the BM cavity of femurs with an α-MEM (Gibco, Life Technologies) supplemented with penicillin (100 U/ml), streptomycin (100 µg/ml), and 10% heat-inactivated FBS. Red blood cells were destroyed in a hemolysis buffer (consisting of 140 mM NH₄Cl and 17 mM Tris-HCl, pH 7.2) for 10 min on ice, and the treated cells were preincubated in a 100-mm dish for 1 h. The nonadherent cells (hematopoietic cells) were harvested and seeded at 50,000 cells per well in 96-well plates in the presence of 20 ng/ml of recombinant human M-CSF (R&D Systems, Inc.) for 2 d. The cells proliferated in response to M-CSF are defined as BMMs in this paper. To generate osteoclasts, the BMMs were further cultured in the presence of 20 ng/ml of M-CSF and 100 ng/ml of human recombinant soluble RANKL (Oriental Yeast Co., Ltd.), with the culture medium changed every 2 d. To maturate macrophages, BMMs were incubated in 20 ng/ml of M-CSF for 6 d with the medium changed every 2 d.

Adherent cells were fixed in 10% formalin for 3 min, followed by further fixation in a fixative buffer consisting of 50% ethanol and 50% acetone for 1 min. The cells were stained with Naphthol AS-MX phosphate (Nacalaitesque) and Fast-Red (Nacalaitesque) for 10–15 min at room temperature and washed with water three times. TRAP-positive cells with >3 or 10 nuclei were counted as multinucleated osteoclasts.

CD14⁺ cells from human peripheral blood were purchased from Lonza. These cells were seeded at 100,000 cells per well in a 96-well plate (Iwaki) and cultured in the presence of recombinant human M-CSF (30 ng/ml; PeproTech) and sRANKL (100 ng/ml) with 1 µg/ml of BSA or NA2-BSA for 8 d. The culture medium was changed every 2 d. These cells were stained with Naphthol AS-MX phosphate and Fast-Red, and TRAP⁺ cells with >50 µm diameter were counted as osteoclasts.

Murine DCIR cDNA was amplified using the following primers: sense 5′-CGG​GAT​CCC​ACC​ATG​GCT​TCA​GAA​ATC​ACT​TAT​G-3′ and antisense 5′-CGG​AAT​TCT​CAT​AAG​TTT​ATT​TTC​TTC​A-3′. The PCR product was cloned into pMXs-IRES-Puro (kindly provided by T. Kitamura, The University of Tokyo, Tokyo, Japan) via the BamHI and EcoRI sites. The pMXs vector was amplified in Escherichia coli (DH5α). DCIR Y/F mutation was generated using a KOD-Plus-Mutagenesis Kit (TOYOBO). In brief, to replace the tyrosine in the ITIM of DCIR to phenylalanine, we designed a forward primer 5′-CAC​TtT​TGC​AGA​AGT​GAA​GTT​CAA​GAA​TGA​ATC-3′ (adenosine was changed into lowercase T) and a reverse primer 5′-ATT​TCT​GAA​GCC​ATG​GTG​GGA​TCC​TTG​GTT​AAC-3′. The inverse PCR was performed using the pMXs vector as a template, which was digested by Dpn I capable of digesting methylated DNA from typical E. coli cell lines. Non-digested PCR products were self-ligated by T4 polynucleotide kinase and ligase at 16°C for 1 h. A Plat-E packaging cell line (provided by T. Kitamura) was cultured in 10% FBS α-MEM and split every second day at a ratio of 1:5. The retroviral vectors were transfected in Plat-E cells by Lipofectamine 2000 (Invitrogen). After 8 h of incubation, the medium was changed and further incubated for 24 h. The retroviral supernatant was collected and frozen at −80°C until it was used. To retrovirally transduce DCIR and DCIR mutant BMs, nonadherent BMCs were cultured in 100 µl of medium containing 20 µl of retroviral supernatant, and the medium was replaced every 2 d.

BMCs underwent blood cell lysis with a hemolysis buffer (consisting of 140 mM NH₄Cl and 17 mM Tris-HCl, pH 7.2) on ice. The BMMs and osteoclasts were detached and washed with a FACS buffer consisting of 2% FBS in PBS to remove serum components. The Fc receptors on these cells were blocked with 2.4G2, an Fc receptor blocker, at 4°C for 10 min. These cells were washed twice with a FACS buffer and incubated with fluorescence-conjugated antibodies for 30 min on ice. Commercially available antibodies are listed in Table S1.

To detect the DCIR ligand, 10 µg/ml of DCIR-Fc-protein, mutated DCIR-Fc-protein, or Fc-protein were preincubated with 5 µg/ml of FITC-conjugated anti-human IgG (Jackson ImmunoResearch Laboratories, Inc.) in a TSA buffer containing 2 mM calcium chloride, 2 mM magnesium chloride, and 0.5% BSA in Tris-buffered saline for 15 min at room temperature. Macrophages and osteoclasts were stained with the protein complex for 1 h at 4°C, followed by gentle washing with the TSA buffer. Flow cytometry was performed on FACSCanto II (Becton Dickinson) or Attune NxT flow cytometer (Thermo Fisher Scientific) and analyzed using FlowJo software (Tree Star, Inc.).

For intracellular cytokine staining, cells were first stained with antibodies against cell surface markers. The cells were fixed and permeabilized using an intracellular fixation and permeabilization buffer set (eBioscience) and then stained for intracellular cytokines, such as IFN-γ, IL-17, TNF-α, GM-CSF, and Ki67.

Total RNA was extracted from BMMs and osteoclasts using the GenEluteTM Mammalian Total RNA Miniprep kit (Sigma-Aldrich). Quantification of total RNA was performed using NanoDrop. 1 µg of total RNA was reverse-transcribed using Superscript II Reverse transcription (Invitrogen) to synthesize first cDNA. The quantitative PCR was performed in a total reaction volume of 10 µl including 5 ng equivalent to total RNA, SYBR Green I (Takara Bio), and a set of forward and reverse primers (Operon) in duplicate or triplicate samples. Quantification was performed on CFX384 Real-Time system (Bio-Rad Laboratories, Inc.). Cycling program was 1 cycle of 95°C for 1 min, followed by 44 cycles of a set of 95°C for 3 s and 60°C for 30 s. The primer sets used to detect each transcript in macrophages or osteoclasts are described in Table S2. The final concentration of the primers was 400 nM. GAPDH was used as an internal housekeeping gene control. The target genes and the housekeeping gene were quantified simultaneously in one plate, with water samples as the negative reference. The difference in RNA quality and initial quantity of samples was normalized to GAPDH. The relative expression of the targeted gene (2^−ΔΔCt method) was calculated from the following equation: 2^−ΔCt, where ΔCt = average Ct for the gene of interest – average Ct for the housekeeping gene. The fold change of a targeted gene on days 3 and 4 relative to a sample on day 2 was determined by the following: 2^−ΔΔCt, where ΔΔCt = (average Ct for gene of interest – average Ct for the gene of a housekeeping gene) each sample on days 3 and 4 – (average Ct for gene of interest – average Ct for the housekeeping gene) sample on day 2. The SD was calculated according to the mathematical method described in earlier reports (Bookout and Mangelsdorf, 2003; Livak and Schmittgen, 2001).

We performed RT-PCR in 20 µl of reaction volume with Ex-taq (Takara Bio). A primer set for amplifying DCIR was sense 5′-CAT​TTC​CCT​TAT​CTC​GCC​CTG​G-3′ and anti-sense 5′-GCA​TGA​GTG​TCC​AAG​ATC​C-3′, which amplified a 686-bp product. The PCR reaction was run in the following program: 30 cycles of 98°C for 10 s, 55°C for 30 s, and 72°C for 1 min. GAPDH was used as an internal control.

Murine DCIR cDNA covering EC from 70 to 238 at the amino acid level was subcloned into pFUSE-hIgG2-Fc2 vector (Invivogen) at the Bgl II site. cDNA coding the EC was generated with a primer set: sense 5′-ATA​AGA​TCT​CAA​AAG​TAC​TCT​CAA​CTT​CTT-3′ and anti-sense 5′-ATA​AGA​TCT​TAA​GTT​TAT​TTT​CTT​CAT​C-3′. Site-directed mutations of the CRD domain in which glutamic acid and serine were replaced with alanine were established by KOD-Plus-Mutagenesis Kit (TOYOBO) according to the manufacturer’s instructions. The primer set for the mutagenesis is as follows: sense 5′-GGT​GCT​CCC​GCT​AGT​GGC​AAT​GAA​AAA​TGT​GCT-3′ and anti-sense 5′-TAG​TGT​AAG​ACC​GTG​TTA​CCA​CGA​GGG​GCG​ATC​A-3′. The sequences encoding human DCIR EC (70–237 amino acids) were amplified by PCR with a primer set (sense 5′-ATA​CCA​TGG​GTC​AAA​AAT​ATT​CTC​AGC​TTC​TT-3′ and anti-sense 5′-ATA​AGA​TCT​TAA​GTG​GAT​CTT​CAT​CAT​C-3′). The resulting PCR fragment was ligated into a pFUSE-hIgG2-Fc2 vector (Invivogen) via NcoI and BglII. We generated three plasmid vectors for DCIR-Fc (the EC of mDCIR), mutant DCIR-Fc (inactive form of CRD domain, E197A/S199A), and Fc-protein (constant region of Ig). We transfected the plasmids into HEK293T cells by Lipofectamine LTX (Invitrogen), followed by incubation in Opti-MEM (Invitrogen) including 2% low IgG FBS (Invitrogen) for 15 d with medium change every 3 d. The Fc-chimeric proteins were purified from collected supernatant by affinity chromatography on Protein A-Sepharose (GE Healthcare).

Murine GnT-II (Mgat2) cDNA from osteoblasts were inserted into pcDNA3.1(+)/FLAG vector (Invitrogen) at BamHI and EcoRI sites. The cDNA was synthesized with a primer set: BamH1, 5′-ATA​GGA​TCC​ATG​AGG​TTC​CGC​ATC​TAC​AAA​CG-3′; EcoRI, 5′-ATA​GAA​TTC​CTG​CAG​TCT​TCT​ATA​ACT​TTT​ACA​GAG​TTC​ATG​G-3′. We transfected the vector into HEK293T cells by Lipofectoamine LTX (Invitrogen) and incubated in Opti-MEM for 24 h. The transient Mgat2-293T cells were harvested with trypsin digestion and resuspended in a TSA buffer for the FACS analysis.

WT and Dcir^−/− BMMs were induced to differentiate osteoclasts with M-CSF and RANKL, as described above. NA2–polyacrylamide (PAA), purchased from GlycoTech, was applied at the beginning of the culture of nonadherent BMCs and consistently added each time the medium was replaced. The number of osteoclasts was determined by TRAP staining.

Primary calvarial cells were enzymatically harvested and seeded at 200,000 cells/well in 12-well plates in α-MEM containing vitamin D3 (10⁻⁸ M) and prostaglandin E2 (10⁻⁶ M) for 24 h. A 10-fold number of nonadherent BMCs was co-cultured in the primary osteoblasts for 7 d, and the medium was changed every 3 d. In some experiments, neuraminidase or β-galactosidase (ProZyme Inc.) was applied in the co-culture system, and the enzymes were added at the time the medium was changed.

Primary osteoblasts cultured for 5 d were treated with either tunicamycin (Sigma-Aldrich; final concentration, 1 µg/ml) or benzyl-α-GalNAc (Sigma-Aldrich; final concentration, 2 mM) for 24 h at 37°C. The osteoblasts were harvested with trypsin and resuspended in a TSA buffer. The BM-derived macrophages for 6-d cultures were harvested using Cell Dissociation Solution Non-enzymatic 1× (Sigma-Aldrich) and resuspended in the TSA. The cells were treated with 5 mU neuraminidases (from Arthrobacter ureafacines; 10269611001; Roche; specific activity is ∼25 U/mg protein) consisting of a mixture of isoenzymes (L, M1, M2, and S) at 37°C for 30 min. These cells were washed with the TSA buffer and then subjected to an FACS analysis.

To measure the effect of the removal of sialic acid on osteoclastogenesis, osteoclasts were induced from nonadherent BMCs in the presence of 0.1 U/ml neuraminidase (Roche). Neuraminidase was replaced when the culture medium was changed every second day. In the co-culture system of osteoblasts with BMCs to induce osteoclasts, 0.1 U/ml of neuraminidase was presented at the start of the cell culture. Neuraminidase was replaced every 3 d.

Calvarial cells were prepared by sequential digestion in accordance with the standard method. Briefly, calvaria from 1- to 2-d-old neonates were surgically isolated, and the adherent mesenchymal tissues were trimmed. The neonatal calvaria underwent a first digestion by a degrading solution containing 0.1% collagenase (Wako) and 0.1% dispase (Roche Diagnostics Japan) for 10 min at 37°C with agitation. The solution was discarded to remove debris, and the remaining calvaria were further digested by the same degrading solution for 1 h at 37°C. The cells isolated from the serial digestions were suspended in an α-MEM containing 10% FCS, penicillin (100 U/ml), and streptomycin (100 µg/ml), and plated at a concentration of 2 × 10⁵ cells/well in 12-well plates. After 2 d in culture, the cells were incubated in an osteogenic medium (α-MEM supplemented with 10% FBS, 50 µg/ml ascorbic acid [Sigma-Aldrich], 10 mM β-glycerophosphate [Calbiochem], and antibiotics) for 14 d or 21 d, with the medium replaced every 3 d for the entire duration.

Calvaria cells were obtained from newborn mice as described in the Materials and methods section. Red blood cells were lysed in a hemolysis buffer for 10 min at 37°C. The remaining cells were resuspended in 90 µl of autoMACS Running Buffer (Miltenyi Biotec Inc.) per 10⁷ cells and incubated with 10 µl of CD11b MicroBeads (Miltenyi Biotec Inc.) per 10⁷ cells for 15 min at 4°C. The cells were washed by autoMACS Running Buffer and resuspended up to 10⁸ cells in 500 µl of autoMACS Running Buffer, and passed through a 75 nylon mesh (75 × 75 µm mesh) to remove cell clumps. CD11b-positive and -negative fractions were separated by an auto MACS Pro separator (Milteny Biotec Inc.), and these cells were used as OsteoMacs or primary osteoblast cells, respectively.

The glycan-protein interaction was detected in a glycan microarray with the evanescent-field fluorescence-assisted detection system developed by Tateno et al. (2008). Briefly, glycan probes, including glycoproteins and glycoside-PAA, were immobilized on microarray-grade epoxy-activated glass slides (Schott AG) in triplicate spots using a noncontact microarray printing robot, MicroSys 4000 (Genomic Solutions Inc.). The glass slides were incubated at 25°C for 3 h and washed with a reaction buffer consisting of 25 mM Tris–HCl, pH 7.4, containing 0.8% NaCl, 1% (vol/vol) Triton X-100, 1 mM MnCl₂, and 1 mM CaCl₂, to remove immobilized materials, followed by blocking in TBS (25 mM Tris-HCl, pH 7.4, with 0.8% NaCl and 1% BSA) at 20°C for 1 h. A protein complex of DCIR-Fc, mutant DCIR-Fc, or Fc with Cy3-labeled anti-human Fc antibodies was formed in a reaction buffer at room temperature for 20 min and kept in the dark. The protein complex was applied to a chamber of the glass slide at 100 μl per chamber and was incubated at 20°C for 3 h. Without washing, the binding was detected by an evanescent-field activated fluorescence scanner, SC-Profiler (Moritex), and the data were analyzed by the Array Pro analyzer Ver. 4.5 (Media Cybernetics, Inc.). Subtracted raw intensity from background intensity is net intensity, which is presented as the average ± SD of triplicate spots.

NA2-BSA, which contains ∼15 NA2 molecules on a BSA molecule, or BSA (10 µg/ml in TBS consisting of 20 mM Tris and 150 mM NaCl, pH 7.4), were precoated on a 96-well plate at 4°C for overnight. Human DCIR-Fc proteins (5 µg/ml) in TBS containing 2 mM CaCl₂ and 2 mM MgCl₂ were incubated with or without 30 mM EDTA at room temperature for 1 h. After washing with TBS, peroxidase-conjugated anti-human IgG antibody (Jackson ImmunoResearch Laboratories, Inc.) was added and incubated at room temperature for 1 h. The binding of Fc-proteins to the NA2 polymers was determined by measuring the absorbance at 450 nm. The NA2 polymer was kindly provided by GlyTech, Inc. (Osaka, Japan).

WT and Dcir^−/− male mice 8–12 wk old were intravenously administered with either neuraminidase (0.1 U/mouse; Roche) or PBS five times every second day. 24 h after the last injection, whole blood samples were collected by cardiac puncture under the anesthesia condition. The serum concentration of calcium ions was determined with the Calcium E-HA test (Wako). The TRAP activity was determined using a TRAPCP & ALP Assay Kit (Takara Bio) by measuring the absorbance at 405 nm.

CIA was induced as previously described (Inglis et al., 2008). Briefly, chicken type II collagen (IIC; Sigma-Aldrich) was dissolved in 10 mM of acetic acid (4 mg/ml) overnight at 4°C and stored at −80°C until use. An emulsion was formed by combining CFA (Difco) with an equal volume of the IIC solution (stocked as 4 mg/ml) for DBA/1J mice. IFA (Thermo Fischer Scientific) supplemented with 1 mg/ml of Mycobacterium tuberculosis H37Ra (Becton Dickinson Co.) was used for the immunization of C57BL/6J. Mice 8–12 wk old were intradermally immunized with 100 µl of emulsion for DBA/1J or 200 µl for C57BL/6J at three distinct sites in the back, near the base of their tails. A total of 14 d after the first immunization, the mice were intradermally immunized with the same amount of CFA/IIC in the vicinity of the first injection site as a boost. The development of arthritis was monitored for 40 d. The arthritis score was evaluated in each limb according to the following criteria, and the maximal score of arthritis severity was 12 per mouse: 0 = normal, 1 = slight swelling and/or erythema, 2 = severe swelling and/or erythema, and 3 = ankylosis of the joint.

WT and Dcir^−/− mice were intravenously administered with 100 µl of neuraminidase (2 U/mouse, cloned from Clostridium perfringens, NEB [P0720]) or PBS 1 d before the first immunization of CFA/IIA. One day after the first immunization, the mice were intravenously injected with either neuraminidase or PBS, followed by six additional administrations, one every 2 d. DBA/1J mice were intravenously administered with neuraminidase available from Roche (0.1 U/mouse) or PBS.

On day 45 after CIA induction, whole blood was obtained by cardiac puncture under anesthesia conditions. The blood samples were left to clot at 4°C for overnight and centrifuged at 3,000 rpm for 10 min to separate serum from coagulated blood. The serum was collected and stored at –80°C until use. IIC (20 µg/ml in PBS) was coated on a 96-well plate at 4°C overnight. The wells were blocked with 10% FBS in PBS at room temperature for 1 h. Serum samples were diluted 5,000-, 2,500-, 500-, 500-, and 100-fold for total IgG, IgG1, IgG2a, IgG2b, and IgG3, respectively. These diluted samples were applied and incubated at room temperature for 2 h. HRP-conjugated goat anti-mouse IgG at 0.8 µg/ml (Jackson ImmunoResearch Laboratories, Inc.) or alkaline phosphatase-conjugated anti-mouse IgG1, IgG2a, IgG2b, and IgG3 at 0.4 µg/ml (Santa Cruz) was added and incubated at room temperature for 1 h. Then, 3, 3′, 5, 5′-tetramethylbenzidine solution (DAKO, Agilent Technologies) for IgG or p-nitrophenyl phosphate solution (Sigma-Aldrich) for IgG1, IgG2a, IgG2b, and IgG3 was added as substrate, and 1 N HCl or 1 N NaOH was used to stop color development. The titers of antibody were determined by measuring the absorbancy of 450 nm for IgG and 405 nm for IgG subclasses.

Sera were prepared as described above, at the end of clinical monitoring of CIA. Using undiluted sera, concentrations of IFN-γ and IL-17 were measured with the mouse ELISA kits (R&D Systems), and the concentration of IL-6 with the mouse ELISA MAX (BioLegend).

CIA was induced in WT and Dcir^−/− mice, or DBA/1J mice with or without neuraminidase injection. At the end of observation, whole blood samples were collected by cardiac puncture under an anesthesia condition. The serum concentration of CTX-I was determined with a CTX-I detection ELISA kit using a competitive assay system, according to the manufacturer’s procedure (Chondrex, Inc.).

Two ankle joints of the hind limbs were amputated on day 45 of CIA. The joints were fixed with 10% neutral formalin and decalcified in 10% EDTA. They were embedded in paraffin blocks, and paraffin sections (4 µm) were prepared for H&E staining. Tissue images were taken with NanoZoomer-SQ (Hamamatsu Photonics K.K.). Arthritic severity scores were calculated for the following four parameters: inflammation, pannus formation, cartilage damage, and bone damage. One section from the right and left ankle joint of the hind limbs was evaluated, and each of these parameters has a total of three points for each joint. The criteria of arthritic severity for each parameter are as follows. For inflammation, 0 = normal, 1 = local infiltration of a few inflammatory cells, 2 = broad local infiltration of inflammatory cells, and 3 = broad infiltration into the joint capsule. For pannus formation, 0 = normal, 1 = pannus formation at less than two sites, 2 = pannus formation at less than four sites, and 3 = pannus formation at more than four sites. One broad pannus formation was counted as two sites. For cartilage damage, 0 = normal, 1 = small loss of articular chondrocytes, 2 = cartilage degradation in one region, and 3 = cartilage degradation in more than two regions. For bone damage, 0 = normal, 1 = rough surface of the talus, 2 = shallow loss of the talus, and 3 = deep loss of the talus. Paraffin sections (4 µm) were stained with Naphthol AS-MX phosphate and Fast-Red for 30 min at room temperature. TRAP-positive cells were counted in sections of the right and left ankle joint containing tibia, talus, calcaneus, navicular, cuneiform, metatarsal bones, and phalanges.

Mice were subcutaneously immunized with 100 µg of pMOG, which was emulsified in 200 µl of IFA (Thermo Fischer Scientific) and PBS (1:1) supplemented with 500 µg nonviable M. tuberculosis H37RA (Difco Laboratories Inc.), on days 0 and 7. The mice were intraperitoneally injected with 200 ng of pertussis toxin (List Biological Labs, Inc.) in PBS on days 0 and 2. Clinical scores of EAE were monitored for 30 d according to the following criteria: 0, no disease; 0.5, partial limp tail; 1, paralyzed tail; 1.5, paralyzed in one hind limb; 2, both hind limbs paralyzed; 2.5, weakness in one forelimb; 3, both forelimbs weakened; 4, forelimbs paralyzed; and 5, moribund. The mean clinical score was determined by averaging the scores of all mice in each group. The cumulative score is the sum of the daily clinical scores of each individual mouse and indicated as an average of each group. The average maximal score and onset are calculated as an average of the maximal clinical score or date of EAE onset of each individual mouse in each group.

WT and Dcir^−/− mice were intravenously administered with 100 µl of neuraminidase (0.1 U/mouse; Roche) or PBS 6 d before the first immunization of pMOG. The mice were intravenously injected with either neuraminidase or PBS, followed by seven further injections administered every 2 d.

Mice administered with mixed anesthetic agents, containing medetomidine, midazolam, and butorphanol, were intracardially perfused with PBS by drip infusion, followed by perfusion with 10% neutral formalin. Spinal cords were dissected and transferred sequentially into 10% sucrose, 20% sucrose, and then 30% sucrose, at 4°C overnight. The tissues were frozen in Tissue-Tek (Sakura Finteck), and frozen sections (6 µm) between L3 and L5 position were prepared using a microtome (Leica CM3050S; Leica Camera AG). Serial frozen sections were stained with H&E or KB staining, and a double staining with Cresyl violet solution and Luxol fast blue (Muto Pure Chemicals Co. Ltd.), and photos were taken with NanoZoomer-SQ. Four tissue sections were used to evaluate the inflammation severity of the spinal cord. The criteria of encephalomyelitic severity are as follows: 0, no inflammation; 1, cellular infiltration only in the perivascular areas and meninges; 2, mild cellular infiltration in parenchyma (inflammatory cells infiltrated less than one third area of the white matter); 3, moderate cellular infiltration in parenchyma (inflammatory cells infiltrated more than one third area of the white matter); and 4, severe cellular infiltration in parenchyma. Demyelination conditions were assessed by KB staining with the following criteria: 0, no demyelination; 1, mild demyelination; 2, moderate demyelination; and 3, severe demyelination (Okuda et al., 2002; Seno et al., 2015).

For the co-culture of 2D2 T cells and DCs, splenic cells were prepared from WT and Dcir^−/− mice, and red blood cells were removed by a hemolysis buffer. Resuspended cells were treated with 2.4G2 at 4°C for 15 min to block Fc receptors. After washing with a FACS buffer, the cells were incubated with 15 µl of CD11b- and CD11c-MicroBeads per 10 million cells for 30 min at 4°C. CD11b- and CD11c-positive fractions were separated with an auto MACS Pro separator (Miltenyi Biotec), and these cells were used as antigen-presenting cells.

Cells from the spleens and the auxiliary, branchial, cervical, inguinal, and mesenteric lymph nodes were treated with a hemolysis buffer to remove red blood cells. The Fc receptors on these cells were blocked with 2.4G2 at 4°C for 15 min, and the cells were incubated with 10 µl of Thy1.2 MicroBeads per 10 million total cells for 30 min at 4°C. Thy1.2-positive fractions were separated with an auto MACS Pro separator, and these cells were used as 2D2 T cells.

2D2 T cells (2 × 10⁵ cells) were cultured either with WT or Dcir^−/− DCs (2 × 10⁴ cells) with or without pMOG (10 µg/ml and 100 µg/ml) for 3 d in a 96-well round-bottom plate in RPMI supplemented with 0.1 mM of β-mercaptoethanol, 100 U/ml of penicillin, 100 µg/ml of streptomycin, and 10% heat-inactivated FBS (the culture medium is called as R10). 2D2 T cells and DCs were sorted with the auto MACS Pro separator described above. Neuraminidase (Roche) was added at the concentration of 0.1 mU/ml at the start of the culture (day 0) and the day after the culture was started (day 1). For the assay of NA2-mediated inhibition, a 96-well plate was coated with 2 µg/ml NA2-BSA or BSA at 4°C for overnight. The plate was washed twice with PBS before it was used. The cells were subjected to the cell surface and intracellular staining for FACS analysis.

For T cell proliferation assay, 2D2 T cells (2 × 10⁵ cells) and DCs (4 × 10⁴ cells) were co-cultured for 3 d with pMOG (10 µg/ml) with or without neuraminidase (0.01 mU/ml; Roche). The cell proliferation was determined with a colorimetric immunoassay based on the incorporation of BrdU, according to the manufacturer’s procedure (Roche). Briefly, 3 d after initiating with the co-culture, BrdU (final concentration was 10 µM) was added, and the cells were incubated for additional 6 h. The cells were fixed, and the DNA was denatured at room temperature. After removing the fixing and denaturation solution, the anti-BrdU-POD antibody was added and reacted for 90 min at room temperature. The BrdU incorporation was determined by measuring the absorbance of 450 nm with a reference wavelength of 570 nm.

BMCs were isolated by flushing the BM cavity of the femurs and tibias with R10. Red blood cells were destroyed in a hemolysis buffer for 10 min on ice. The BMCs were cultured at the concentration of 2 × 10⁶ cells in 10 ml R10 in a nontreated 100-mm dish with 20 ng/ml of GM-CSF (Peprotech). 3 d after incubation, 10 ml of R10 containing 20 ng/ml of GM-CSF was added to the dish. At 6 and 8 d, 10 ml of supernatant was collected, and the cell pellet, after centrifugation at 1,500 rpm for 5 min, was resuspended with an equal volume of R10 containing 20 ng/ml of GM-CSF, followed by the restoration of the suspension into the dish. At 10 d, cells in the supernatant were used as GM-DCs.

Thy1.2⁺ cells from spleens and lymph nodes were sorted with an auto MACS Pro separator (Miltenyi Biotec). Anti-CD3 antibody (10 µg/ml; Biolegend) was coated on a 96-well plate at 4°C for overnight. The Thy1.2⁺ cells were seeded at 2 × 10⁵ cells per well with soluble anti-CD28 antibody (2 µg/ml) and cultured at 37°C for 3 d. The supernatants were stored before used for the detection of cytokines by ELISA. For the staining of intracellular molecules, cells were treated with 2.4G2 (Fc blocker) on ice for 30 min. These cells were permeabilized with fixation/permeabilization solution (BD Biosciences) on ice for 30 min and stained with antibodies against GM-CSF, IFN-γ, IL-17, and Ki67, and analyzed by FACS.

CD11c⁺ cells were isolated from the co-culture system with 2D2 T cells in the presence of pMOG (100 µg/ml) for 3 d using an autoMACS proseparator and were exposed to CellCover (ANACYTE Laboratories) to protect RNA. The purity of the cell population was 95% or more in the FACS analysis. RNA samples were extracted using NucleoSpin RNA XS (Takara Bio), and the RNA quality was assessed by RNA Screen Tape in TapeStation System (Agilent Technologies) to ensure an RNA integrity number >7.0. RNA-seq library construction followed by next-generation sequencing were conducted by GENEWIZ. Briefly, the total RNA was quantified by a Qubit RNA Assay (Invitrogen) and qualified in the TapeStation system. 25 ng of total RNA was used for mRNA enrichment of poly(A) selection with NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs). mRNA samples were fragmented and reverse-transcribed for the first strand and second strand cDNA using random primers. The 5′ and/or 3′ ends of the fragmented RNA were repaired, and indexed adaptors were added via ligation. RNA-seq libraries were prepared using a MGIEasy RNA Directional Library Prep Set V2.0 (MGI Tech) according to the manufacturer’s instructions. The generated libraries were quantified by Qubit DNA Assay, and the fragment size was assayed by D1000 DNA Screen Tape in TapeStation. Libraries with unique sample barcodes were pooled and sequenced using DNBSEQ-G400 (MGI Tech) in a 150-bp paired-end configuration. Approximately 20 million paired-end reads per sample were yielded, and nearly 90% of the sequence had a Q30 score.

The raw sequencing data were processed using TrimGalore v0.6.6 to remove adapter sequences and low-quality reads. The processed reads were used to align to the mouse genome reference mm10 using hisat2 v2.2.0. Aligned read counts that fell on the exons of each gene were calculated by the R package Rsubread v1.32.4. The read counts were normalized and compared for differential gene expression using R package edgeR v3.24.3. The RNA-seq data have been deposited in the Gene Expression Omnibus database under accession no. GSE184068.

A GO enrichment analysis was performed by the PANTHER Overrepresentation Test (Released 20200728) for the analysis of functional pathways. Differentially expressed genes that were fold-change >0 and P value <0.05 were selected for the GO enrichment analysis. The functional signatures of gene sets were recategorized as below. Genes annotated with the MHCII protein complex, antigen processing and presentation, and the regulation of T cell proliferation were described as antigen presentation and T cell proliferation. Genes annotated with a pattern recognition receptor signaling pathway and a transmembrane receptor protein tyrosine kinase signaling pathway were described as receptor signaling pathways.

The statistical significance between groups was determined using the two-tailed unpaired Student’s t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001). The clinical and histological scores of CIA and EAE were evaluated using the Mann–Whitney U test or one-way ANOVA with Dunnett’s post hoc test, and the incidence of CIA and EAE was analyzed using the χ² test. The difference between the groups was considered statistically significant at P < 0.05.

Fig. S1 shows the binding of DCIR to N-linked glycans on macrophages, osteoclasts, and osteoblasts in a CRD- and Ca²⁺-dependent manner. Fig. S2 provides the data of DCIR-Fc or Fc-protein binding on a glycan array and shows DCIR expression in macrophages and OsteoMacs. Fig. S3 shows the effect of intravenous administration of neuraminidase on DCIR-Fc binding to BMCs. In addition, the production of anti-collagen antibody is comparable in CIA-induced mice with neuraminidase treatment. Fig. S4 provides flow cytometry data of cytokine production from T cells stimulated with pMOG in neuraminidase treatment. Fig. S5 shows cytokine production from T cells activated with anti-CD3/CD28 antibody after neuraminidase treatment. Table S1 lists the antibodies. Table S2 shows PCR primer sets.

We thank A. Akitsu, K. Shimizu, and A. Tobinai-Sugawara for their helpful discussion and technical support.

This work was supported in part by a grant for the Core Research for Evolutionary Science and Technology from the Japan Science and Technology Agency (105100000222), a Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (S; 24220011) and a Grant-in-Aid for Scientific Research on Innovative Areas (20H04954), a grant from the Japan Agency for Medical Research and Development (16809407) to Y. Iwakura, and a Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (C; 23500489) to T. Kaifu.

Author contributions: T. Kaifu and R. Yabe conducted most of the experiments, R. Yabe prepared protein materials and performed ligand screening, and T. Kaifu analyzed DCIR signaling mechanisms. T. Kaifu, T. Maruhashi, and N. Fujikado analyzed bone morphometric data; T. Kaifu and S.H. Chung assessed an animal model of CIA. J. Hirabayashi and H. Tateno carried out glycan array screening; T. Kaifu wrote the draft manuscript. Y. Iwakura organized and supervised the project and edited the draft manuscript. All authors discussed the data and approved the final manuscript.