Cell wall N-glycan of Candida albicans ameliorates early hyper- and late hypo-immunoreactivity in sepsis

Severe infection often causes a septic cytokine storm followed by immune exhaustion/paralysis. Not surprisingly, many pathogens are equipped with various anti-inflammatory mechanisms. Such mechanisms might be leveraged clinically to control septic cytokine storms. Here we show that N-glycan from pathogenic C. albicans ameliorates mouse sepsis through immunosuppressive cytokine IL-10. In a sepsis model using lipopolysaccharide (LPS), injection of the N-glycan upregulated serum IL-10, and suppressed pro-inflammatory IL-1β, TNF-α and IFN-γ. The N-glycan also improved the survival of mice challenged by LPS. Analyses of structurally defined N-glycans from several yeast strains revealed that the mannose core is key to the upregulation of IL-10. Knocking out the C-type lectin Dectin-2 abrogated the N-glycan-mediated IL-10 augmentation. Furthermore, C. albicans N-glycan ameliorated immune exhaustion/immune paralysis after acute inflammation. Our results suggest a strategy where the immunosuppressive mechanism of one pathogen can be applied to attenuate a severe inflammation/cytokine storm caused by another pathogen. Kawakita et al use a murine sepsis model to show that N-glycan from pathogenic C. albicans ameliorates sepsis - and thus improves survival - through upregulation of the cytokine IL-10. This study demonstrates a potential strategy in which the immunosuppressive mechanism of one pathogen can be applied to attenuate a severe inflammation caused by another pathogen.

I nfection leading to sepsis continues to be a major health concern 1,2 . It is estimated that 30 million people worldwide are affected each year. In sepsis, microbes and viruses cause severe inflammation accompanied by an excess production of proinflammatory cytokines, commonly referred to as a cytokine storm, resulting in multi-organ failure 3 . In many cases, after the inflammation is resolved, immune paralysis is induced, leading to the majority of septic deaths 1,2 . Therefore, practical methods to control these septic responses are urgently required.
IL-10 is a potent immunosuppressive cytokine that plays pleiotropic roles in the management of the immune system 4,5 . Various types of immune cells produce IL-10, and, in turn, suppress immune system activity by downregulating the production of proinflammatory cytokines, including IL-1β, TNF-α, and IFN-γ, which, when overproduced, generate the cytokine storm. IL-10 suppresses acute systemic inflammation, as in sepsis 6 , as well as T cell exhaustion 7 that leads to hypoimmunoreactivity and recurrent/further infections. Furthermore, IL-10 also ameliorates acute lung inflammation and injury associated with influenza infection 8 . These properties make IL-10 and the IL-10 biosynthetic pathway attractive candidates for treatments of various types of diseases, such as colitis, arthritis and pancreatitis 9 . In principle, control of IL-10 expression in vivo could be extended to the treatment of severe inflammation and inflammatory diseases. Disappointingly, however, several trials of injection of exogenous IL-10 showed low efficacy and various side effects 5 , partly due to site and timing of its actions. As an alternative, induction of IL-10 in vivo could possibly improve these problems.
In the 1970's, an immunosuppressive factor was found in the blood of candidiasis patients 10 . The factor comprised mannoproteins released from the cell wall of Candida albicans 10,11 , and it suppressed candida-induced proliferation of lymphocytes and delayed-type hypersensitivity (DTH) of the patients 10 . These effects were eliminated from patients by treatments with antimycotic agents 10 . In the mannoproteins, the polysaccharide mannan is indispensable for the immunosuppressive activities 12 . Thus, it is conceivable that the C. albicans evasion of the host immune system principally involves the immunosuppressive mannoprotein. However, identification of the effective mannan structure conferring the immunosuppressive activities remained elusive, primarily due to technical difficulties, e.g., chemical synthesis of polysaccharides.
In this report, we focused on C. albicans mannoprotein, for which immunosuppressive activities have been documented in candidiasis patients. Mannoprotein-derived N-glycan, purified from clinically isolated C. albicans strain J-1012, was tested in a mouse model where sepsis was induced by LPS. In this model system, N-glycan administration ameliorated early and late septic responses. Furthermore, the accumulation of structural information of N-glycans from several Candida/yeast strains enabled us to identify mannan structures that are indispensable for the activities.

Results
Amelioration of LPS-induced septic responses by C. albicans N-glycan. N-glycan was purified from mannoprotein of C. albicans strain J-1012 by Fehling's method 13 (hereinafter referred to as J-1012 N-glycan). Administration by i.v. injection of J-1012 Nglycan (400 μg N-glycan/20 g mouse weight) along with a low dose of LPS (15 μg/20 g mouse weight) resulted in up-regulation of IL-10 and down-regulation of proinflammatory IL-1β, TNF-α, and IFN-γ in serum, compared to administration of LPS alone (Fig. 1a). J-1012 N-glycan did not affect LPS-induced IL-6 and MCP-1production. The J-1012 N-glycan also augmented IL-10 production in vivo in response to a TLR2 ligand, Pam 3 CSK 4 , but showed up-regulation of TNF-α and IL-6 production (Supplementary Fig. 1). In response to LPS, J-1012 N-glycan alone induced small amounts of these cytokines. IL-10 production was not observed in TLR4 knockout (KO) mice, consistent with essential role of TLR4 in LPS recognition (Fig. 1b). This also indicates that the IL-10 production is driven by LPS rather than a possible contaminant in J-1012 N-glycan. β-elimination, which minimizes contamination of J-1012 N-glycan by O-glycan, did not affect the induction of IL-10 ( Fig. 1c). Composition analysis of J-1012 N-glycan by gas chromatography detected contamination by small amounts (~3%) of glucose, possibly derived from βglucan in the cell wall. However, lyticase treatment, which degrades β1,3-glucan, did not affect IL-10 induction (Fig. 1d). Furthermore, mice lacking the β-glucan receptor Dectin-1 produced comparable levels of IL-10 to those of WT mice in response to J-1012 N-glycan and LPS (Fig. 1e). To assess the correlation between IL-10 and the other cytokines, we co-injected neutralizing anti-IL-10 mAb along with J-1012 N-glycan and LPS. As a result, large amounts of TNF-α and IFN-γ were observed, consistent with the control of these cytokines by IL-10 ( Fig. 1f). To test the efficacy of J-1012 N-glycan in a more challenging context, a lethal dose of LPS (100 μg/20 g mouse weight) was administered, and co-injection of mice with J-1012 N-glycan not only augmented serum IL-10 production (Fig. 1g), but improved survival (Fig. 1h). Neutralization of IL-10 by mAb negated the improvement (Fig. 1i). Taken together, these results indicate that J-1012 N-glycan augments production of IL-10 in vivo. In turn, elevated levels of IL-10 could suppress the cytokine storm, leading to a reduction of septic death.
Involvement of α-mannan and mannose core of the N-glycan. We previously reported that J-1012 N-glycan contains various types of side chains (Fig. 2a), including α-mannan side chains (55.4% of total side chains), β-mannan-capped α-mannan side chains (41.9%) and phosphomannan side chains (2.7%) 13 . Specific structural patterns in the side chains may be involved in the augmentation of IL-10 production. To investigate this possibility, we first treated J-1012 N-glycan with α-mannosidase to remove α-mannan side chains, phosphomannan and mannose/ Man 10 GlcNAc 2 core (designated as Man 10 Core), but not βmannan-capped α-mannan side chains (Fig. 2a). This treatment of J-1012 N-glycan reduced the IL-10 production and restored the TNF-α and IFN-γ induction by LPS in vivo (Fig. 2b), suggesting that there are structural features of J-1012 N-glycan that are sensitive to α-mannosidase. Moreover, C. albicans NIH B-792 N-glycan without β-mannan-capped α-mannan side chains showed the augmentation of IL-10 ( Supplementary Fig. 2), implying the dispensability of the structure. We next tested N-glycan purified from S. cerevisiae 4484-24-D1 (mnn1/mnn4), which contains α1,2-mannan side chains, but not phosphomannan. This N-glycan induced larger amounts of IL-10 than J-1012 N-glycan did (Fig. 2c), underscoring the essential contribution of α-mannan side chains and/or Man 10 Core. This result also suggests that β-mannose-capped side chains in J-1012 N-glycan suppresses the IL-10 augmentation. We then tested N-glycan from C. lusitaniae (non-albicans strain), in which the side chains are predominantly β-mannan-capped α-mannan (77.5%) rather than α-mannan (14.9%) and phosphomannan (7.6%) 14 . Indeed, this N-glycan elicited significantly less IL-10 induction in vivo than J-1012 N-glycan (Fig. 2d). This result revealed the importance of α/β-mannan composition in the side chains. Following this line of reasoning, it is possible that N-glycan containing long α-mannan side chains without βmannan-capped α-mannan side chains should efficiently induce IL-10. As expected, N-glycans isolated from C. stellatoidea (Candida albicans (Robin) Berkhout) and C. parapsilosis (nonalbicans strain) (Fig. 2a) significantly induced large amounts of IL-10 in comparison with J-1012 N-glycan in the sepsis model (Fig. 2e). Both N-glycans showed TNF-α and IFN-γ levels equivalent to those induced by J-1012 N-glycan. On the other hand, C. albicans NIH B-792 N-glycan, whose structure is similar to that of C. stellatoidea, but with phosphomannan side chains consisting of β-mannan, showed the same extent of IL-10 augmentation as J-1012 N-glycan ( Supplementary Fig. 2), suggesting that there are negative effects of β-mannan on IL-10 augmentation. These results indicated that the α-mannan structure in the side chains affects the augmentation of IL-10 production. However, N-glycan from S. cerevisiae X2180-1A-5 (mnn2) 15 , which consisted exclusively of an α1,6-mannan backbone and Man 10 -Core still showed IL-10 augmentation in vivo (Fig. 2f), suggesting that there are effective contributions from 2 different linkages, as well as the α-mannan structure in the side chains.
To find more clues about the mannan structure required for IL-10 induction in vivo, we identified the host receptor for J-1012 N-glycan. We first sought to identify the cell type involved in the IL-10 production using clodronate-liposomes that eliminated phagocytic cells in vivo ( Supplementary Fig. 3). This treatment significantly decreased IL-10 production in response to J-1012 Nglycan and LPS (Fig. 2g). In vivo treated resident peritoneal macrophages (rpMϕ) showed augmentation of IL-10 production by J-1012 ( Supplementary Fig. 4). In vitro, purified rpMϕ grown on plates coated with J-1012 N-glycan produced large amounts of IL-10, but not TNF-α, after stimulation with LPS (Fig. 2h). The production of IL-10 was sensitive to treatment with NaIO 4 -NaBH 4 , which damages the saccharide structure ( Fig. 2i), or αmannosidase, which digests α-mannan in the J-1012 N-glycan ( Fig. 2j). NaBH 4 reducing aldehyde group generated by NaIO 4 did not affect the IL-10 production alone. On phagocytic cells, there are several lectins that recognize yeast mannose/mannan 16 . To further narrow down the list of putative receptors for J-1012 N-glycan, the effects of various types of signaling inhibitors were tested, using the rpMϕ assay ( Fig. 2k and Supplementary Fig. 5). Of these inhibitors, GW5074, a Raf-1 kinase inhibitor that suppresses IL-10 production via human DC-SIGN 17 , showed no effect on the IL-10 induction. In addition, SIGNR1, a homolog of human DC-SIGN, causing oral tolerance through IL-10 induction in intestine 18 , was dispensable for the augmentation of IL-10 production by J-1012 N-glycan in vitro and in vivo (Fig. 2l). On the other hand, piceatannol, a Syk kinase inhibitor, suppressed the IL-10 induction effectively (Fig. 2k). Of the lectins using Syk signaling, Dectin-2 is known to recognize α-mannan of C. albicans 19 . We therefore checked Dectin-2KO mice, and observed markedly impaired IL-10 induction by J-1012 N-glycan in vivo (Fig. 2m). Consistent with this observation, Dectin-2KO mice were significantly more susceptible to a lethal dose of LPS with J-1012 N-glycan (Fig. 2n). Furthermore, IL-10 augmentation by N-glycan from S. cerevisiae X2180-1A-5 (mnn2) also depended on Dectin-2 (Fig. 2o). Taking into consideration the fact that  Dectin-2 does not recognize α-1,6 mannan 20 in the backbone, these results suggest that Dectin-2 recognizes the mannose core of J-1012 N-glycan, leading to augmentation of IL-10 production in vivo.
Cytokine induction by synthetic mannose core though Dectin-2. We next synthesized a Man 9 Core conjugated with BSA (Man 9 Core-BSA) and tested its ability to induce IL-10 in the rpMϕ system. Man 9 Core-BSA coated plates induced production of IL-10 and the other cytokines (Fig. 3a), while control conjugated Man 3 Core-BSA and mannose-BSA (Man 51 -BSA), which is known to induce oral tolerance through SIGNR1 18 , did not (Fig. 3b, c). Unexpectedly, however, the response of rpMϕ to Man 9 Core-BSA was independent of Dectin-2 (Fig. 3d). Moreover, the response of rpMϕ to J-1012 N-glycan depended on Dectin-2, but only to some extent (Fig. 3e). These results suggest that other receptors using Syk signaling, e.g., Mincle (Supplementary Fig. 6) and Dectin-3 16 , whose expression levels are higher in Mϕ than DCs, are involved in Man 9 Core recognition in vitro. On the other hand, the response of bone marrow-derived DCs (BMDCs) to Man 9 Core-BSA depended on Dectin-2 ( Fig. 3f) and on CARD9, a member of the signalosome for Syk 21 (Supplementary Fig. 7). BMDCs did not respond to Man 3 Core-BSA and Man 51 -BSA (Fig. 3g), similar to the lack of response in rpMϕ. The IL-10 production in response to Man 9 Core-BSA was down-regulated by J-1012 N-glycan in the medium, consistent with the notion that Fig. 3 Augmentation of IL-10 production by mannose core in vitro. a-e Cytokine production by rpMϕ from WT(a-c) and Dectin-2KO (d, e) mice after stimulation with Man 9 Core-BSA, Man 3 Core-BSA, Man 51 -BSA (bound more than 51 mannose monomers), and J-1012 N-glycan, immobilized on plates as in Fig. 2h. Schematic structures of Man 9 Core-BSA and Man 3 Core-BSA are depicted. f IL-10 production by BMDCs from WT and Dectin-2KO mice in response to Man 9 Core-BSA. g IL-10 production by BMDCs from WT mice upon stimulation with Man 9 Core-BSA, Man 3 Core-BSA, Man 51 -BSA, J-1012 N-glycan, and J-1012 N-glycan treated with α-mannosidase. h Schematic structure of (Man 9 CoreCys) 5 -biotin. εAHX, ε-amino-n-hexanoic acid. i Cytokine production by BMDCs from WT and Dectin-2KO mice upon stimulation with (Man 9 CoreCys) 5 -biotin bound to an avidin-coated plate. These experiments were repeated at least twice, and representative results are shown. Data are expressed as the mean ± SD. p value was determined by unpaired two-tailed Student's t-test. N.S., not significant (p > 0.05).
there is steric hindrance between the mannose core and the Nglycan ( Supplementary Fig. 8) Furthermore, biotinylated penta-Man 9 Core ((Man 9 CoreCys) 5 -biotin), high-density Man 9 Core with a minimum carrier (Fig. 3h), stimulated IL-10 production by BMDCs in a Dectin-2-dependent manner (Fig. 3i). In addition, the (Man 9 CoreCys) 5 -biotin also induced TNF-α, MCP-1 and IL-6 production from BMDCs in a Dectin-2-dependent manner. These results with synthetic glycans further implicate Dectin-2 in the recognition of the mannose core in the N-glycan. Collectively, our results suggest that the mannose core of C. albicans N-glycan is recognized by Dectin-2, leading to the IL-10 augmentation and subsequent suppression of the early onset of septic responses in vivo.
Amelioration of hypo-immunoreactivity by C. albicans N-glycan. Hypo-immunoreactivity/immune paralysis is often observed after several days or weeks of sepsis, leading to recurrent and opportunistic infections followed by septic death in the majority of cases 1 , and to an increased risk of cancer development in humans 22 . The secondary infection is also the case with severe virus infection 23 . At present, there is no practical treatment for the hypoimmunoreactivity. It is also possible that the IL-10 augmentation by J-1012 N-glycan affects the hypo-immunoreactivity after the resolution of acute inflammation. We therefore investigated the effects of the J-1012 N-glycan on the hypo-immunoreactivity response. First, DTH response was evaluated. DTH is caused by antigen-specific effector T cells pre-sensitized. After challenge with the same antigen, the T cells induce productions of cytokines including IFN-γ and trigger type IV hypersensitivity, e.g., skin swelling. We first observed that i.v. LPS injection immediately after sensitization with sheep red blood cells (SRBC) suppressed the DTH response in footpad after 1 week, indicating hypo-immunoreactivity after sepsis induction (Fig. 4a). This weak response was reversed by co-injection of LPS with the J-1012 N-glycan concomitant with the SRBC sensitization, suggesting improvement of the hypo-immunoreactivity, instead of aggravation. We next observed the effects of the J-1012 N-glycan on antigen-specific IFN-γ production several weeks after sepsis induction, because pre-clinical adjunctive immunotherapy involving IFN-γ seems to improve the hypoimmunoreactivity 24 . After 2 weeks of sepsis induction in the presence of ovalbumin (OVA) as an antigen, cytokine production of splenocytes was analyzed in vitro, showing that the J-1012 N-glycan enhanced antigen-specific IFN-γ production, which possibly supported a Th1 reaction in DTH 25 ; notably, however, IL-10, TNF-α, and IL-6 production were not enhanced (Fig. 4b). The elevated IFNγ levels were also observed 1 month after sepsis induction (Supplementary Fig. 9). The IFN-γ enhancement after 2 weeks was prevented by co-injection of neutralizing mAb to IL-10 along with J-1012 N-glycan and LPS (Fig. 4c), suggesting that the initial augmentation of IL-10 by J-1012 N-glycan was involved in the IFN-γ production after alleviating the inflammation. We next checked the effects of J-1012 N-glycan on T cell phenotypes after septic inflammation. After transfer of transgenic DO11.10 T cells, sepsis was induced with OVA and J-1012 N-glycan. After 2 weeks, DO11.10 T cells were analyzed. In spleen, the effects of J-1012 N-glycan on the number of DO11.10 T cells were minimal (Supplementary Fig. 10a, b). However, the expression levels of various types of surface proteins relative to T cell functions were altered; we observed an increased CD62L + and a reduced PD-1 high population in comparison with LPS alone (Fig. 4d and Supplementary Fig. 10c). [27][28][29] . On the other hand, expression of Dectin-2 on T cells in lymph node and spleen was undetectable before or after stimulation with i.v. injection of LPS ( Supplementary Fig. 11), suggesting that there is no direct influence of the N-glycan on T cells. Moreover, 2 weeks after sepsis  6). a-c Data are expressed as the mean ± SD and are representative of at least two independent experiments (n = 3). p value was determined by unpaired two-tailed Student's t-test. N.S., not significant (p > 0.05). R.IgG, control rat IgG. SRBC sheep red blood cells.

CD62L is involved in homing of T cells into lymph nodes to induce immune responses, and PD-1, an immune checkpoint receptor, negatively regulates T cell activation. PD-1 is observed on exhausted T cells 26 and its expression is upregulated in septic patients
induction, MHC class II expression and the M1/M2 balance of Mϕ in spleen seemed identical in the presence or absence of J-1012 N-glycan ( Supplementary Fig. 12). Taken together, it is possible that J-1012 N-glycan improves the late hypo-immunoreactivity by protecting T cells from excessive activation through the early augmentation of IL-10 production.

Discussion
Our approach, based on a putative immune suppressing mechanism of C. albicans, points to its potential to control IL-10 production leading to amelioration of both the early hyper-and late hypo-immunoreactivity in sepsis. Clinically, sepsis treatment has to manage both phases. Nowadays, extensive care enables us to suppress the early hyper-inflammation. However, the suppression doesn't improve the late hypo-immunoreactivity. As a result, the hypo-immunoreactivity has become the major cause of septic death. To avoid the hypo-immunoreactivity, some cytokines, e.g., IFN-γ, are used to maintain immunocompetence. However, it is difficult to use the cytokines because there are few markers to monitor the septic phase, in hyper or hypo-immunoreactivity, of the patient. From this point of view, the bidirectional properties of the N-glycan may lead to methods for biphasic sepsis treatment. On the other hand, the current understanding of sepsis is based on a balance between a systemic inflammatory response syndrome (SIRS) and a compensatory anti-inflammatory response syndrome (CARS). It is believed that CARS caused by IL-10 and the other immunosuppressive cytokines induces immune paralysis. However, at least in our system, large amounts of IL-10 in early sepsis did not induce immune paralysis. Instead, IL-10 augmented by N-glycan maintained immunoreactivity, suggesting that the origin and local environment, rather than the total amount of IL-10 production, affects subsequent immune competence after alleviation of the acute inflammation. From this point of view, the α-mannans in the N-glycan are expected to directly act on DCs, a center of the immune system, through Dectin-2. This may be why the N-glycan can ameliorate the septic cytokine storm and maintain T cell reactivity in vivo. Our future studies will move forward on artificial neo-glycan conjugates including mannose core and α-mannan side chains, and on testing their effects on acute inflammation caused by microbial and viral infection.
We observed in vivo IL-10 augmentation by N-glycans from several non-albicans candida strains and S. cerevisiae strains as well as three C. albicans strains, J-1012, NIH B-792, and C. stellatoidea. Therefore, the IL-10 augmentation depends on the structure of N-glycans rather than yeast strains. In the N-glycans, the composition of the side chains may affect their IL-10 augmentation capacity. Our results suggest that α-mannan and βmannose-capped α-mannan side chains positively affect and inhibit IL-10 augmentation by the mannan core, respectively. These results are informative for the design of artificial neoglycan conjugates.
The glycans likely will act in humans because the mannoprotein was originally found to be immunosuppressive in candidiasis patients. Dectin-2 may be a keystone to control both the hyperand hypo-immunoreactivity at once. Dectin-2 is also a pharmaceutical target to manipulate IL-10 production in vivo, and to treat chronic and refractory candidiasis. At present, we do not know how intracellular Dectin-2-Syk signaling enhances cytokine production by LPS. Co-operation with the other receptors is also possible. Furthermore, stimulation of Dectin-2 by mannose core causes not only IL-10 production, but also the other cytokines' productions in vitro. It will be important to identify the major IL-10-and IFN-γ-producing cells, and to establish how Dectin-2 is involved in the augmentation of IL-10 production and suppression of the other cytokines in vivo.

Ethics statement
All animal experiments were approved by Animal Research Committee, Graduate School of Biostudies, Kyoto University (protocol number Lif-K20008).
The 2 h at room temperature to reduce the end generated. After washing the plate, purified rpMϕ were seeded and stimulated as described above.
Induction of endotoxin shock and cytokine production from splenocytes. For low-dose endotoxin shock, mice were injected i.v. with HANKS' balanced salt solution (HBSS) (200 μl/20 g mouse weight) containing 75 μg/ml of LPS or 75 μg/ml of Pam 3 CSK 4 with or without 2 mg/ml of N-glycan from C. albicans J-1012 and other strains. Serum cytokines were analyzed as described below. In some cases, mice were simultaneously injected i.v. with OVA (200 μg/20 g mouse weight), anti-IL-10 mAb (50 μg/20 g mouse weight) and control rat IgG. For analyses of in vitro cytokine production, after 2 weeks or as indicated, splenocytes (7.5 × 10 5 ) were stimulated with LPS (100 ng/ml) or OVA (100 μg/ml) for 4 days in IMDM containing 10% FCS, antibiotics and 50 μM of β-mercaptoethanol. Cytokine production was analyzed as described below. For lethal endotoxin shock, mice were injected i.p. with HBSS (200 μl/20 g mouse weight) containing 500 μg/ml of LPS with or without 2 mg/ml of J-1012 N-glycan.
Analyses of cytokine production. Cytokines in serum/culture supernatant were assessed by a Cytometric Bead Array mouse inflammatory kit and Mouse IL-1β Flex Set (BD Biosciences, Franklin Lakes, NJ) in accordance with the manufacturers' protocols using a FACSCalibur system and an Accuri C6 Plus (BD Biosciences).
Treatment of mice with clodronate liposomes. To deplete phagocytic cells in vivo, mice were treated i.v. and i.p. with anionic clodronate liposomes (For-muMax Scientific, CA) (60 μl/20 g mouse weight, respectively. Total 120 μl). After 24 h, elimination of phagocytic cells in spleen and peritoneal cavity was analyzed by flowcytometry, and sepsis was induced with the aforementioned low dose of LPS.
Preparation and stimulation of BMDCs. Bone marrow-derived DCs (BMDCs) were prepared from Balb/c bone marrow cells unless otherwise indicated 35 . Briefly, bone marrow was collected from femurs of adult mice and cultured in RPMI containing 10% FCS and 250 U/ml GM-CSF (gifted from Kirin Brewery). After 3 day in culture, equal volume of new media was added. On 6 day and 8 day, half of culture medium was replaced with new medium. Cells were harvested on 12 day. For CARD9KO BMDCs, bone marrow cells (C57BL/6 background) were kind gifts from Dr. Hiromitsu Hara (Kagoshima University, Kagoshima, Japan). BMDCs were purified using anti-CD11c-magnetic beads (Miltenyi Biotech, Bergisch Gladbach, Germany). For cytokine production analyses, BMDCs (1 × 10 5 cells) were stimulated with LPS (100 ng/ml) for 24 h, in some cases, in the presence of J-1012 N-glycan (100 μg/ml) in medium. Biotinylated (Man 9 CoreCys) 5 was immobilized on an avidin-coated culture plate (Thermo Scientific, MA) for 12 h at 10 μg/ml.
Analyses of delayed-type hypersensitivity (DTH). For DTH analyses, mice were immunized by injection of 2 × 10 6 SRBC in 25 μl of HBSS into the left hind footpad and subsequently treated i.v. with HBSS (200 μl/20 g mouse weight) containing 75 μg/ml of LPS with or without 2 mg/ml of J-1012 N-glycan. After 7 days, the mice were given 1 × 10 8 SRBC in 25 μl of HBSS into the right hind footpad. The footpad swelling was measured with a dial gauge caliper (Mitutoyo Corporation, Kanagawa, Japan) before and 24 h after antigen challenge.
Statistical analysis. Data are expressed as the means ± SD of triplicate assays. Statistical significance was determined by the two-tailed Student's t-test. Differences in the survival of each group were determined by the Wilcoxon test. All experiments were performed twice or more times and representative results are shown. In some experiments, compiled results are shown.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All raw data is stored and securely backed-up and available upon request. All data underlying the graphs are available as Supplementary Data 1-5. Other data are available on reasonable request from the corresponding author.