Allergy-Protective Arabinogalactan Modulates Human Dendritic Cells via C-Type Lectins and Inhibition of NF-κB Free

The impact of allergic diseases is rising unimpeded in industrial countries. The hygiene hypothesis proposes a causal explanation for this phenomenon, namely a reduction in infectious disease in early childhood (1). Although a tempting explanation, epidemiological studies deliver inconsistent results (2–4). In the search for other conditions that are protective against allergic disease, recent studies have shown that growing up and living in rural areas decrease the risk of developing allergies significantly compared with living in cities (5). In particular, children, when they were exposed to a traditional farming environment in the rural areas of the Alps, were protected in this way. Further studies showed that contact with barns in the earliest childhood resulted in allergy-protective effects (6–9).

These allergy-protective effects have been traced back to different factors, such as the inhalation of cowshed dust extracts (CDE) or single components isolated from CDE (10–13).

One allergy-protective molecule that makes up 13% of CDE total mass was identified as arabinogalactan (AG) (14). It was shown in a mouse asthma model that AG isolated from Alopecurus pratensis can prevent allergic airway inflammation and sensitization. AGs are polysaccharides that are ubiquitously present in most plants and even in mycobacteria, but not in animals. They consist predominantly of galactose and arabinose. Further sugars, such as rhamnose, fucose, and mannose, occur to a lesser extent. Different AG preparations are already in use in several traditional and natural medical products worldwide and are subject to studies due to their immune-modulating properties (15–17).

Polysaccharides mediate their immune-modulatory properties in mammals predominantly via binding to carbohydrate receptors. Dendritic cells (DCs) express a wide pattern of these receptors with a variety of different functions. Among these receptors, the group of C-type lectin receptors (CLRs) is the largest one. The DC-specific ICAM-3–grabbing nonintegrin (DC-SIGN) and the macrophage mannose receptor 1 (MMR-1) are two commonly expressed CLRs on DCs. Both receptors are known to interact with mannose and fucose, but they differ in their number of carbohydrate recognition domains and cellular signaling. Several studies have identified exogenous ligands binding to DC-SIGN, such as the HIV, measles virus, Mycobacterium tuberculosis, and other pathogens. Furthermore, many pathogens use DC-SIGN not only to enter the DC, but also to inhibit DC function (18, 19). Unlike the members of the group of TLRs, these CLRs induce cellular activation, only to a much lesser extent. However, stimulation of these receptors, such as DC-SIGN, may modulate TLR-induced cytokine secretion and DC maturation.

We have previously shown that AG isolated from CDE, hay, or different grasses is able to protect from allergic disease in a mouse model of asthma. Therefore, it is of great interest whether AG is also able to modulate the human immune response. Additionally, mice, compared with humans, have five different DC-SIGN homologs named SIGNR1–SIGNR5, but there are contrary findings as to which of them is the functional homolog to human DC-SIGN (20–22). In fact, these homologs have neither the same ligand specificity nor the same signaling pathways compared with human DC-SIGN. Besides DC-SIGN, there are even more structural and functional differences in CLRs between mice and humans (23). Therefore, we chose to investigate the effect of AG in a human model of T cell activation to get direct insight into the modulation of the human immune system by AG.

One major problem concerning in vitro studies with substances isolated from plants is the contamination with pathogen-associated molecular patterns (PAMPs). In particular, when working with cells of the innate immune system, tiny amounts of PAMPs can have a major influence on the results of the experiments. Therefore, we have focused in the present study on the production of contamination-free AG from plant cell tissue cultures. We aimed to investigate whether this PAMP-free AG modifies the immune-stimulatory capacity of human monocyte-derived DCs (moDCs). Moreover, we sought to identify receptors to which AG bind and to get insights into the molecular mechanisms behind immune modulation.

For flow cytometric analysis, anti–CD1a-PE (clone HI149; Immunotools, Friesoythe, Germany), anti-CD14 (clone TLK4; Miltenyi Biotec, Bergisch Gladbach, Germany), anti-CD209 (clone DCN47.5; Miltenyi Biotec), and anti-CD83 (clone HB15a; Beckman Coulter, Krefeld, Germany) were used. Anti-human phospho–NF-κB p65 (S529) PE (clone B33B4WP) and CFSE were purchased from eBioscience (San Diego, CA). Recombinant human (rh)DC-SIGN–Fc chimera and rhMMR-1 were purchased from R&D Systems (Minneapolis, MN). Anti-human IgG1-biotin Ab and sandwich ELISA for human IFN-γ, IL-6, IL-12p70, IL-10, and TNF-α were purchased from BD Biosciences (Erembodegem, Belgium). Blocking Abs against DC-SIGN (clone DCN47.5) and human MMR (clone 15-2) as well as an murine IgG1 isotype control Ab (clone T8E5) were purchased from InvivoGen (Toulouse, France). The concentration of Ab used for blocking experiments was 10 μg/ml. These Abs were carefully selected to not take influence on the activation of cells by LPS (Supplemental Fig. 1). Anti–His-biotin Ab was purchased from Qiagen (Hilden, Germany). The β-d-glucosyl Yariv reagent [1,3,5-tris-(4-β-glucopyranosyl-oxyphenylazo)-2,4,6-trihydroxybenzene] was prepared according to Yariv et al. (24). p-Aminophenyl-β-d-glucoside (1.8 mM) was diazotized with sodium nitrite (1.8 mM) in 10.8 ml 0.5 M HCl at 0°C and added dropwise to 0.45 mM phloroglucin dissolved in 22.5 ml water, with the pH being kept at 9 by addition of 0.5 M NaOH. After 2 h the pH became stable and an equal volume of methanol was added. The precipitate was collected by centrifugation (4000 × g) and dissolved in 40 ml water. The precipitation step was done again and the residue was dried at 40°C. The Yariv reagent was stored frozen at −20°C as a 1 mg/ml solution in 0.9% sodium chloride. rhGM-CSF and rhIL-4 were purchased from Immunotools.

The surface of P. pratense (Poaceae) seeds were sterilized by treatment for 10 min in 5% thymol solubilized in 96% ethanol, followed by 10 min in 2% sodium hypochloride. After washing with sterile water, the seeds were grown on plates prepared with 1% agarose in Murashige–Skoog medium containing sucrose and the following phytohormones: 2 μg/ml 2,4-dichlorophenoxyacetic acid, 4 μg/ml α-naphthaleneacetic acid, and 0.3 μg/ml kinetin (all purchased from Sigma-Aldrich, St. Louis, MO). After germination at 27°C in darkness, parts of the shoot were cut and transferred to fresh agar plates to generate callus tissues. After 2 wk of growing, callus tissue was transferred to Murashige–Skoog liquid medium containing the same phytohormone concentrations as described above for the agar plates. Suspension culture was performed in the absence of antibiotics. Suspension-cultured cells were then incubated at 27°C at 120 U/min in a horizontal shaker. Supernatants were collected after centrifugation (10 min, 3000 × g).

Supernatants of suspension cultures were then pooled and autoclaved. In this step proteins and peptides were denaturized so that the mucus-like structure was resolved. To remove insoluble components, three filtration steps were included. At first, pooled and autoclaved callus supernatants were filtered with a paper filter, afterward a glass fiber filter, and at last a 0.22-μm cellulose filter was used. The raw AG-containing solution was concentrated by using a 30-kDa filtration unit (Merck Millipore, Billerica, MA). AG concentration was determined by β-d-glucosyl Yariv radial diffusion gel as described previously (14). For precipitation of AG, an appropriate amount of Yariv’s reagent was added to the raw AG-containing solution. After centrifugation for 1 h at 21,000 × g, sediment was washed once with sterile 0.9% NaCl solution. Afterward, 1 ml saturated sodium dithionite solution was added to the sediment and incubated at 60°C until the red color disappeared completely. AG was then purified by size exclusion chromatography on a PD-10 column (GE Healthcare, Fairfield, CT) and eluted in 1-ml fractions. AG-containing fractions were identified by radial diffusion assay and pooled. Concentration was then determined by a resorcin assay and radial diffusion gel with AG from gum arabic (Biosupplies Australia, Melbourne, VIC, Australia) as a standard. The substance was proved to be LPS free by TLR4 assay as described previously (25). Molecular mass of the preparation was determined by size exclusion chromatography as described by Göllner et al. with following modifications: column set (two PL aquagel-OH 40, 8 μm, 300 × 7.5 mm and one PL aquagel-OH mixed, 8 μm, 300 × 7.5 mm) and flow rate (0.7 ml/min) (26).

Membranes of human moDCs were prepared as follows. Briefly, DCs were washed with ice-cold PBS and resuspended in 20 mM HEPES buffer with 250 mM sucrose, 1% digitonin, and protease inhibitor mixture (Roche, Basel, Switzerland). Cells (2.4 × 10⁷) were mechanically homogenized and centrifuged for 7 min at 700 × g and 4°C. Membrane-containing supernatant was collected and membranes were sedimented by centrifugation for 60 min at 16,500 × g and 4°C and washed twice. Membranes were then resuspended in 1 ml cold HBSS containing protease inhibitor mixture (Roche, Basel, Switzerland). Resuspended membranes (100 μl) with a protein concentration of 1.6 mg/ml were then incubated with 100 μg AG for 1 h at 4°C in HBSS buffer to allow receptors to bind AG. Afterward, AG-receptor complexes were precipitated by adding β-d-glucosyl Yariv’s reagent. Precipitates were washed twice with ice-cold HBSS, and afterward reducing SDS-PAGE loading buffer (27) was added. Sample bands from the SDS-PAGE were further analyzed by mass spectrometry.

For proteomics experiments, liquid chromatography–tandem mass spectrometry analyses of tryptically digested proteins were carried out on an UltiMate 3000R SLC nano system online coupled to an Orbitrap Elite instrument (both Thermo Scientific, Bremen, Germany). Separated gel bands obtained after membrane receptor precipitation assay were digested in gel with trypsin. Samples of human moDCs were digested in solution after lysis and protein solubilization with RapiGest (Waters, Eschborn, Germany) following the manufacturer’s instructions.

For protein identification, Proteome Discoverer software (ver. 1.3, Thermo Scientific, Waltham, MA) was used to search against UnipPotKB/Swiss-Prot database with Mascot search algorithm. Taxonomy was restricted to humans. Peptide modifications were considered at methionine (oxidation [variable]) and cysteine (propionamide for in-gel digestion [variable] and carbamidomethyl for in-solution digestion [fixed]). Precursor mass tolerance was set to 5 ppm, fragment mass tolerance to 0.4 Da, and maximum number of missed cleavage sites to 1. Peptide identifications with >1% false discovery rate were discarded. Quantitative analyses were performed with Progenesis liquid chromatography–tandem mass spectrometry software (ver. 4.0.4265.42984, Nonlinear Dynamics, Newcastle upon Tyne, U.K.). A detailed description of the single steps of quantification has been recently published (28, 29).

To confirm the binding between AG and the identified receptors, an inhibitory ELISA was used. Briefly mannan from Saccharomyces cerevisiae was coated to 96-well ELISA plates (Thermo Fisher Scientific, Waltham, MA). The receptor of interest (DC-SIGN or MMR-1) was preincubated either with dilutions of mannan or AG for 1 h in HBSS buffer at 37°C. Afterward, the preincubation mixtures were transferred to mannan-coated plates and incubated for 1 h at room temperature while shaking. Bound receptors were detected by anti-human IgG1 secondary Ab (DC-SIGN) or anti-His Ab (MMR-1). Detection of biotinylated secondary Ab was carried out by streptavidin-HRP (Sigma-Aldrich) and OptEIA tetramethylbenzidine substrate (BD Biosciences). Absorbance was measured at 450 nm.

Monocytes were isolated from a total of 18 different donors (61% female). Voluntary, healthy donors were selected independently of their sensitization status to allergens. DCs were generated from human blood monocytes as described elsewhere (30, 31). On day 6, DCs were positive for the following surface molecules CD1a (93 ± 1%), DC-SIGN (86 ± 3%), CD83 (15 ± 3%), and CD86 (17 ± 1%). All experimental procedures were approved by the Ethics Committee at Ruhr University Bochum (Bochum, Germany).

moDCs were harvested and stimulated with 1 μM AG. After 4 h of stimulation, 1 ng/ml LPS was added. After 24 h, supernatants were collected and cytokines were measured by ELISA. DCs were harvested, stained for surface markers, and analyzed by flow cytometry.

To detect NF-κB subunit p65 and its phosphorylation, an instant ELISA kit of eBioscience (Affymetrix eBioscience, Frankfurt, Germany) was used according to the manufacturer’s protocol. DCs were stimulated as described above with the difference that activation with LPS was shortened to 1 h before lysis of cells. Intracellular staining of phosphorylated p65 was performed by staining cells that were fixed with paraformaldehyde followed by permeabilization with methanol. Cells were stained with 0.03 μg anti-human phospho–NF-κB p65-PE. Stained cells were analyzed by flow cytometry (CyFlow SL; Partec, Münster, Germany).

Human moDCs were harvested and stimulated with AG, LPS, or both for 24 h. Afterward, cells were washed, collected, and used for comparative quantitative proteomics analysis. In this study, five different monocyte donors were used to obtain statistically relevant data. Afterward, data were analyzed by a “protein information resource” database search for processes associated to the significantly regulated proteins.

For visualization of the proteomic data, proteomaps were generated as described previously (32).

Human naive CD4⁺ T cells were isolated with a human naive CD4⁺ isolation kit II (Miltenyi Biotec). T cells were stained with CFSE, as described elsewhere for proliferation assay (10). moDCs and naive CD4⁺ cells were cocultivated for 6 d at a ratio of 10,000 DCs to 20,000 naive CD4⁺ T cells in 96-well round bottom plates. Cocultured cells were stimulated with AG alone or in combination with LPS. After 6 d of cocultivation, CFSE-stained cocultures were used to determine T cell proliferation by flow cytometry, and supernatants were used to measure T cell cytokines (IFN-γ, IL-5, IL-13, and IL-17A).

All data were analyzed by a Mann–Whitney test, Wilcoxon test, or one-way ANOVA as indicated. GraphPad Prism software (ver. 5; GraphPad Software, La Jolla, CA) or Progenesis liquid chromatography–mass spectrometry (Nonlinear USA, Durham, NC) were used for the analysis. The p values < 0.05 were considered statistically significant. Results are presented as medians or means as indicated.

We isolated AG from P. pratense cell suspension culture. The molecular mass was determined by multiangle laser light scattering to be 127 kDa. AG was tested to be free of TLR2 and TLR4 ligands via TLR-transfected HEK293 cells. After confirming the allergy-protective activity of AG in a mouse model of allergic airway inflammation (Supplemental Fig. 2) we aimed to test whether this molecule may also modulate the behavior of human DCs in vitro.

Therefore we generated moDCs and stimulated the immature cells with 1 μM AG, 1 ng/ml LPS, or both for 24 h to investigate the effects of AG on DC biology. Afterward, we investigated the surface marker expression by flow cytometry, cytokine secretion by ELISA, and cell viability by MTT assay. We observed that AG does not have an influence on CD1a (Fig. 1A), CD14, and DC-SIGN (data not shown) expression, showing that AG does not impair DC generation. Moreover, we found no influence on DC viability, as measured by the MTT test (data not shown). In contrast, stimulation with AG significantly reduced the expression of the activation marker CD83 when it was given 4 h before LPS stimulation (Fig. 1B), showing that AG attenuates the activation of DCs by LPS. CD86 expression seems to be suppressed in a similar manner (data not shown).

Stimulation of DCs with AG leads to a significant increase in IL-6 and IL-10, but not TNF-α or IL-12p70 secretion. LPS, in contrast, induces a much higher amount of cytokine secretion in this setting. The costimulation of DCs with AG and LPS shows that AG significantly reduces LPS-induced IL-6, IL12p70, IL-10, and TNF-α (Fig. 1C–F) secretion. Stimulation of DCs with AG or costimulation with LPS did not show any influence on the IL-1β secretion (data not shown).

In search of receptors to which AG binds on DCs, we used the ability of β-d-glucosyl Yariv reagent to precipitate AG-receptor complexes. Precipitated receptor complexes were fractionated by SDS-PAGE (see Supplemental Fig. 3A) and protein bands were identified by mass spectrometry.

MMR-1, DC-SIGN, and SIGLEC-1 were identified in three independent experiments as receptors with a high probability of binding AG. We focused our further studies on DC-SIGN and MMR, which are both known to modulate the behavior of DCs (33, 34). We used an inhibitory ELISA with AG as the inhibitor to further confirm the binding of AG to DC-SIGN and MMR-1. The binding of DC-SIGN (Fig. 2A) and MMR-1 (Fig. 2A) to coated mannan was inhibited dose-dependently by preincubation with AG. Representative binding curves are shown in Supplemental Fig. 3B for DC-SIGN and for MMR-1 in Supplemental Fig. 3C.

We calculated IC₅₀ values for mannan and AG. Both receptors showed a comparable IC₅₀ for mannan IC_50(DC-SIGN) of 0.28 and IC_50(MMR-1) of 0.12. DC-SIGN showed a significantly lower IC₅₀ for AG IC_50(DC-SIGN) of 0.01 compared with mannan.

To test whether binding of AG to DC-SIGN or MMR-1 explains the altered activation profile found after stimulation of cells with LPS, we repeated the stimulation experiment in the presence of receptor-blocking Abs. Interestingly, blocking of binding of AG to either MMR-1 or DC-SIGN by neutralizing Abs reconstituted production of IL-6 partially (Fig. 3), showing that both receptors seem to be involved in immune modulation.

To get an overview of the expression of cellular proteins that may be regulated by AG, a quantitative proteome analysis was performed. The stimulation of moDCs with AG, LPS, or both showed a specific pattern of differentially regulated proteins (Fig. 4A, Supplemental Table I) linked to distinct cellular processes (Fig. 4B). When DCs were stimulated with AG, these cells showed reduced response to LPS, and therefore we focused our attention on processes that are involved in the negative regulation of TLR signaling. One of these proteins is the E3 ubiquitin-protein ligase tripartite motif–containing protein 21 (TRIM21) that is involved in negative regulation of the NF-κB pathway. TRIM21 was upregulated 2.2-fold after costimulation of cells with LPS and AG (Fig. 5A).

A second protein that is differentially regulated by stimulation with AG, LPS, or the combination of both is the IL-1R antagonist (IL-1RA), which is a secreted and immune regulatory protein. Expression of IL-1RA was upregulated after stimulation with AG (2.7-fold) or LPS (9.5-fold) and the combination of both (15.2-fold) (Fig. 5B).

Owing to the influence of AG on TRIM21 and the reduced expression of proinflammatory cytokines, we measured the expression of NF-κB subunit p65 (Fig. 5C) and its phosphorylation by a semiquantitative ELISA (Fig. 5D). The results reveal that the amount of the p65 subunit in cells that were costimulated with AG and LPS is only half the amount in cells that were stimulated with the different factors alone. Moreover, the amount of the phosphorylated p65 is greatly reduced in cells that were costimulated in comparison with cells that were stimulated with LPS only. The reduced amount of the phosphorylated protein was confirmed by intracellular staining of phosphorylated p65 and subsequent analysis of fluorescence by flow cytometry (Fig. 5E).

We set up an allogeneic MLR to test whether the influence of AG on the biology of DCs has an impact on their T cell stimulatory capacity. This was performed in the presence of AG, LPS, or both.

Stimulation of DCs with LPS induced proliferation of naive T cells in allogeneic MLRs. In combination with AG, however, the T cell proliferation was reduced to baseline level (Fig. 6A). We further investigated cytokine production by the measurement of cell culture supernatants by ELISA. We found that stimulation with LPS induced IFN-γ and IL-17A production, whereas it reduced IL-13 production. Interestingly, simultaneous stimulation with AG was able to reduce LPS-induced IFN-γ and IL-17A secretion, whereas it had no influence on IL-13 (Fig. 6B, 6C, 6E). Production of IL-5 is not affected by stimulation with LPS or by AG (Fig. 6D).

Recently, we showed that treatment of mice with AG isolated from CDE prevents allergic airway inflammation, hyperreactivity, and sensitization (14). The allergy-protective role of AG seems to be mediated by attenuating the Th2-inducing properties of DCs, as shown by sensitization of mice with allergen-pulsed AG-treated bone marrow–derived DCs. One major problem in all studies using substances isolated from the environment is contamination with PAMPs. In particular, cell culture models applying cells of the innate immune system are strongly affected by PAMPs, such as LPS, influencing the results when present at concentrations near the detection limit. Furthermore, LPS itself was associated with allergy-protective activity by inducing LPS tolerance (8). Therefore, one aim of the present study was to produce contamination-free AG from P. pratense suspension culture. The use of P. pratense culture resulted from practical considerations, that is, it grew faster than the tissue of A. pratensis and was therefore a more efficient producer of AG. The efficient production of AG would be one major factor for a possible medical use of AG. We have shown previously that AG from these two related plants that belong to the same family is structurally similar (14).

The first step of allergic sensitization involves the uptake and processing of allergens by DCs that, subsequently, induce a Th2 response. DCs undoubtedly play an important role in the initiation of the allergic immune response in mice as well as in humans. We have chosen human moDCs to study the effects of AG on DC biology because working with these cells as a model system is well established and there are strong experimental data from murine models that this cell type is crucial for orchestrating the asthmatic response (33). At first we sought to identify the receptor to which AG binds. For this purpose, we prepared membranes from human moDCs, incubated them with AG, and precipitated AG-receptor complexes with Yariv’s reagent. Subsequently, we identified the receptors precipitated by SDS-PAGE and mass spectrometry. DC-SIGN and MMR-1 were identified as interesting receptors with immune-modulating properties that were already described in the literature (35). However, the method used here for the identification of the receptors is limited to the fact that membranes are not completely solubilized, resulting in the release of small vesicles containing more than one sort of protein. This implies that receptors identified in this way may be more in the close vicinity of an AG-binding receptor, but do not bind to AG per se. It is possible that Siglec-1 was identified in this artificial way because it is unlikely that it is binding AG owing to the lack of any detectable sialic acid. Therefore, we confirmed binding of AG to DC-SIGN and MMR-1 in vitro with recombinant receptors. A question that arises from this observation is: why do these plant polysaccharides bind to mammal pattern recognition receptors? One possible explanation for this would be the similarity between AG from plants and microbial polysaccharides in the sense of “molecular mimicry.” For instance, mycobacteria do contain AG and lipoarabinomannan polysaccharides in their cell wall (36). It is known that the cell walls of mycobacteria contain a variety of ligands of DC-SIGN (37); however, it needs to be defined whether AG is one of them.

Activation of DC-SIGN is associated with immune regulation. There are several reports showing that DC-SIGN is targeted by a diversity of pathogens, such as M. tuberculosis, Helicobacter pylori, and HIV-1. This is discussed as an immune escape mechanism of these pathogens (38, 39). These reports describe that stimulation of DC-SIGN acts in a regulatory manner via enhancing IL-10 secretion of DCs by different intracellular signal transduction events depending on the ligand. This may lead to DCs with a lower expression of costimulatory molecules exhibiting a lower T cell stimulatory capacity in an allogeneic MLR with CD4⁺ T cells (34). Ligation of MMR-1 is also linked to immune regulation. In some reports, activation of MMR-1 on human moDCs resulted in IL-10 secretion and reduced T cell stimulatory activity of DCs (40, 41). As expected from these publications, we found a significant increase in IL-10 production of cells that were stimulated with AG alone. Besides IL-10, IL-6 was also increased in the supernatant of cells that were stimulated with AG alone. IL-6 was previously found to be increased in DCs that were stimulated with DC-SIGN via a pathway that involved signaling via RAF-1 and ERK (42).

In contrast, we found decreased production of IL-10 in cells that were costimulated with AG and LPS compared with cells that were stimulated with LPS only. Likewise, other cytokines, for example, IL-6, which is known to be involved in the induction of Th2 cells, are reduced due to costimulation of cells with AG. To test whether this reduction in cytokine production is due to binding of AG to MMR-1 or DC-SIGN, we have blocked binding by using neutralizing Abs. Indeed LPS-induced IL-6 production was partially restored after blocking either one of them, showing that stimulation of these CLRs is involved in AG-mediated downregulation of cytokine production. At first glance, the AG-induced IL-6 and IL-10 production contradicts its suppression upon LPS stimulation. However, this may reflect a state of DC-SIGN–induced tolerance in DCs, which, for example, is already known for stimulation of DCs with gp120 (19). This glycoprotein from HIV also leads to low-level production of cytokines such as IL-10 and IL-12 followed by a reduced responsiveness of the cells to LPS.

Therefore, we studied alterations in the proteome of AG-stimulated DCs in search of mechanisms for the downregulation of DC function. We revealed that AG leads to a distinct pattern of regulated processes in a comparative study utilizing a label-free quantitative proteomics approach. We show that costimulation with AG and LPS, but not one substance alone, increases E3 ubiquitin-protein ligase TRIM21 expression in moDCs. Moreover, we found that expression of the p65 subunit of NF-κB is reduced in moDCs that were treated with AG and LPS simultaneously. These findings strongly suggest that costimulation with both substances leads to the increased expression of TRIM21, which then leads to the degradation of NF-κB. It has already been shown that TRIM21 (also known as RO52) leads to the monoubiquitination of IKKβ, and in this manner it leads to decreased NF-κB phosphorylation and activity (43).

Our findings suggest that one effect of AG is the activation of DC-SIGN and MMR-1 resulting in attenuation of DC activation by danger signals. The reduced expression of costimulatory molecules supports this assumption. Reduced expression of costimulatory molecules and proinflammatory cytokines may lead to a lower T cell stimulatory capacity of DCs, thereby resulting in a reduced T cell response in vitro. Indeed, we observed a reduced T cell stimulatory capacity of AG-stimulated DCs in allogeneic MLRs, resulting in reduced proliferation of T cells and reduced production of IFN-γ. Importantly, note that the MLR used in this study is limited to the fact that a Th1 response is induced due to the strong allogeneic stimulus. Although the MLR is a good model to study the activation of T cells leading to proliferation, it is not possible to study T cell polarization without the use of further manipulating reagents, such as IFN-γ–neutralizing Abs. Therefore, the effect on the induction of the Th2 response in humans remains elusive. The upregulation of IL-1RA that was observed by stimulating DCs with AG is also of interest when thinking about a potential mode of action of AG. In fact, different studies were able to show that stimulation with IL-1RA is able to dampen allergic response in mice and inhibit T cell proliferation (44, 45). Polymorphisms in IL-1RA are also associated with asthma and other allergic diseases (46).

In conclusion, we found that AG binds to the human receptors DC-SIGN and MMR-1 on the surface of DCs. Costimulation of DCs with AG and LPS leads to decreased NF-κB p65 subunit expression and phosphorylation, which is likely the reason for reduced DC activation. Furthermore, we showed that AG-treated DCs have a reduced T cell stimulatory capacity. These observations and the fact that AG lacks proinflammatory properties make AG an interesting candidate for the development of allergy-preventing compounds.

We thank Petra Bonowitz and Sandra Busse at Experimental Pneumology, Ruhr University Bochum, for excellent technical assistance. We also thank Philip Saunders (Berlin, Germany) for proofreading the manuscript.

M.P. is a scientific consultant for Protectimmun GmbH. A.B. is an associate of Protectimmun GmbH but has no other commercial relationship to this company. The other authors have no financial conflicts of interest.