Author Correction: Conjugation of wildtype and hypoallergenic mugwort allergen Art v 1 to flagellin induces IL-10-DC and suppresses allergen-specific TH2-responses in vivo

A correction to this article has been published and is linked from the HTML version of this paper. The error has been fixed in the paper.

While the introduction of vaccines has strongly improved our treatment options for many infectious diseases, the search for efficacious vaccines to treat allergic diseases remains a major task. Currently, the only established therapy option for the treatment of allergies is conventional allergen immunotherapy (AIT) with allergen extracts. However, AIT is not convenient for patients due to a multi-year treatment regimen, for some allergies only partially efficacious, and can be hampered by severe allergic side effects 13,14 . To improve AIT, novel vaccine candidates and accompanying adjuvants that increase efficacy while decreasing unwanted adverse-effects are needed 15 . Strategies currently investigated for improved treatment of allergic diseases include for example the usage of different adjuvants, more thoroughly defined recombinant allergens or hypoallergenic allergen-derivatives and allergen-derived peptides combining preserved T cell reactivity with the reduced capacity to activate B cells 16 .
Among the different adjuvants currently tested TLR-ligands are promising candidates to modulate allergen-specific TH2 responses because of their intrinsic ability to induce robust, mostly TH1-biased, immune responses. In this context, the TLR5-ligand flagellin is of special interest, since its proteinous nature enables the generation of fusion proteins consisting of flagellin and antigen using recombinant DNA technologies. Flagellin is a well-known bacterial motility protein with an established adjuvant potential [17][18][19] . In theory, fusion proteins combining antigen and flagellin (as adjuvant) into a single molecule enable efficient targeting of antigens to TLR5 positive APCs as well as the processing and presentation of the fused antigen in the context of the flagellin-mediated APC activation 20 . Indeed, a plethora of studies suggest that flagellin-containing fusion proteins efficiently induce protective humoral and cell mediated immune responses against the conjugated (often otherwise poorly immunogenic) antigen and therefore have potential as vaccine candidates for many different types of infectious diseases [21][22][23][24][25][26][27][28] and allergies 20,29 . In previous studies, we have shown that application of a fusion protein consisting of flagellin A (FlaA) derived from Listeria and the model allergen ovalbumin resulted in the generation of IL-10 producing tolerogenic dendritic cells and was able to prevent the establishment of intestinal allergy in an experimental mouse model 20,29 .
Therefore, the aim of this study was to generate novel vaccine candidates consisting of the TLR5 ligand and the major mugwort allergen Art v 1 and investigate their immune modulating capacity both, in vitro and in vivo. In order to further improve the safety profile of such fusion proteins, we generated two different fusion proteins: rFlaA:Artv1 incorporating the wild type Art v 1 sequence and rFlaA:Artv1 hyp incorporating a hypoallergenic Art v 1 sequence in which seven out of eight cysteine residues were exchanged to serine residues in order to disrupt conformational IgE epitopes 30 . Of note, upon incorporation of the hypoallergenic Art v 1 sequence, the resulting fusion protein retained its capacity to activate, both murine and human dendritic cells and to suppress Art v 1-induced TH2 cytokine secretion ex vivo and in vivo. Therefore, we think that incorporating hypoallergens into flagellin:allergen fusion proteins is a viable strategy to further improve the safety profile of these promising vaccine candidates.
Using the described methods for protein expression and purification all recombinant proteins used in this study were generated with high purity and displayed the expected apparent molecular mass (rArt v 1: 24 kDa, rArt v 1 hyp : 24 kDa, rFlaA: 30.6 kDa, rFlaA:Artv1: 52 kDa, and rFlaA:Artv1 hyp : 52 kDa, Fig. 1A). Identity of the fusion proteins was confirmed by tandem mass spectrometry after tryptic digests (data not shown). IgE-immunoblot analysis using sera of mugwort pollen-allergic individuals showed that IgE-reactivity with Art v 1 was retained for rFlaA:Artv1, whereas no IgE reactivity was observed for rFlaA and the hypoallergenic derivatives rArt v 1 hyp and rFlaA:Artv1 hyp (Fig. 1B). Analysis of secondary structure elements by circular dichroism (CD) spectroscopy revealed both ω-shaped curve progressions and minima typical for mainly α-helical proteins for rFlaA, rFlaA:Artv1, and rFlaA:Artv1 hyp whereas rArt v 1 and rArt v 1 hyp displayed u-shaped curves typical of random coil structures (Fig. 1C). The circular dichroism spectra obtained for rArt v 1 and rArt v 1 hyp are in accordance with previous reports describing dominating random coil structures of rArt v 1 likely caused by its high proline content 3,30 . For further protein characterization, hydrodynamic radii (R H ) of the different proteins were assessed by dynamic light scattering analysis (Fig. 1D). While for rArt v 1 (R H = 2.44 nm), rArt v 1 hyp (R H = 2.56 nm), and rFlaA:Artv1 hyp (R H = 4.46 nm) single radii were observed, for rFlaA (R H = 17.82 + 39.08 + 49.24 nm), and rFlaA:Artv1 (R H = 25.78 + 39.08 + 519.6 nm) multiple fractions with considerably higher hydrodynamic radii were detected, suggesting the formation of high molecular aggregates.
In line with the pronounced secretion of pro-and anti-inflammatory cytokines, a strong upregulation of the maturation and costimulatory molecules CD40 and CD69 was detected on CD11b + CD11c + B220 − mDCs (BALB/c) stimulated with rFlaA:Artv1 or rFlaA:Artv1 hyp (and to a lesser extend the equimolar mixture of both proteins rFlaA + Artv1) (Fig. 2B). Determination of mDC viability upon stimulation with the different proteins showed less than 2% dead cells in both unstimulated cells and cells stimulated with the different proteins ( Supplementary Fig. S2), indicating that in the concentrations used none of the tested proteins had a toxic effect on the tested mDC. In summary, compared to the mixture of non-fused components, stimulation with both fusion proteins strongly promoted mDC activation and led to increased production of pro-and anti-inflammatory cytokines. rFlaA:Artv1 and rFlaA:Artv1 hyp suppress TH2-cytokine production in a co-culture of mDC and Art v 1-specific TH2-biased T cells. Next we investigated the impact of rFlaA:Artv1-and rFlaA:Art-v1 hyp -mediated activation of mDCs on the induction of adaptive immune responses and their ability to modulate allergen-specific T cell responses. We performed co-culture experiments using mDCs in combination with ex vivo isolated, TH2-biased CD4 + T cells from rArt v 1 + Alum sensitized animals and checked whether both fusion proteins were able to modulate TH2-responses induced by Art v 1-re-stimulation in these co-cultures (Fig. 3). Whereas IL-13 (a typical marker cytokine of respiratory allergy) was induced upon re-stimulation of mDC:TC co-cultures with rArt v 1, co-administration of rFlaA:Artv1 or rFlaA:Artv1 hyp , but not with both proteins provided separately or as a mixture, was able to dose dependently and significantly suppress rArt v 1-induced IL-13 secretion (Fig. 3). Interestingly, this suppression of TH2 cytokine secretion was paralleled by high levels of IFN-γ secretion in co-cultures stimulated with rFlaA:Artv1, whereas anti-inflammatory IL-10 secretion was induced by rFlaA:Artv1 hyp (20 µg/ml) (Fig. 3). Here, IL-10 secretion was associated with a suppression of IL-2 secretion (a marker of T cell activation and proliferation) (Fig. 3).
Similar results were obtained from DC:TC co-cultures stimulated with the different proteins in the absence of additional re-stimulation with rArt v 1 ( Supplementary Fig. S3). Interestingly, rFlaA:Artv1 hyp -induced IL-10 and IFN-γ secretion were inversely regulated. Of note, stimulation with both Art v 1 fusion proteins resulted in considerably lower levels of IL-2 and IL-13 secretion from the rArt v 1-specific TH2-biased T cells compared to its unfused counterpart ( Supplementary Fig. S3). Accordingly performed stimulations of mDCs alone in the absence of CD4 + T cells showed, that the cytokines IL-13, IFN-y, and IL-2 detected in the corresponding co-cultures were only produced in mDC:T cell co-cultures whereas the observed IL-10 secretion could be attributed to the mDCs within the mDC:CD4 T cell co-cultures (data not shown).
In summary, both fusion proteins were able to efficiently suppress TH2 cytokine secretion from TH2-biased T cells in the presence of their specific antigen.
Art v 1-induced IL-13 secretion is suppressed after vaccination with rFlaA:Artv1 and rFlaA:Art-v1 hyp . In order to study the immune modulating capacity of both fusion proteins in vivo, we investigated if prophylactic vaccination with the different proteins was able to modulate TH2-biased T cell responses induced by subsequent Art v 1 + Alum sensitization (Fig. 4A). CD4 + T cells were isolated two weeks after the final sensitization, co-cultured with BALB/c mDCs, and checked for cytokine profile upon rArt v 1 re-stimulation (Fig. 4B).
Prophylactic vaccination with rFlaA:Artv1 and rFlaA:Artv1 hyp (and to a lesser extent rFlaA) was able to suppress the rArt v 1-mediated induction of IL-13 and IL-2 secretion from mDC:TC co-cultures (Fig. 4B). Moreover, rArt v 1-induced IL-5 secretion was suppressed after vaccination with both fusion proteins (Supplementary   Fig. S4). Moreover, in this co-culture system generally no IL-4 secretion was detectable, probably due to its low stability in the applied cell culture medium (data not shown). Interestingly, for co-cultures of T cells isolated from animals vaccinated with rFlaA:Artv1 the suppression of cytokine secretion was paralleled by a highly significant increase in IFN-γ levels which was not observed with T cells isolated from animals vaccinated with rFlaA or rFlaA:Artv1 hyp . Here, vaccination with neither rArt v 1 nor the mixture of rFlaA + rArt v 1 did result in a comparable suppression of IL-13 and IL-2 secretion (Fig. 4B). Moreover, in vitro re-stimulation with rArt v 1 alone did not result in high levels of IL-10 secretion (Fig. 4B).
In summary, these results show that prophylactic vaccination with both fusion proteins was able to efficiently suppress the differentiation of TH2 cells in vivo.

Both Art v 1 fusion proteins efficiently induce Art v 1-specific IgG production. Vaccination with
both fusion proteins induced significantly increased levels of rArt v 1-specific IgG1 and IgG2a antibodies compared to animals vaccinated with the respective controls (Fig. 5). Here, significantly lower but still detectable levels of rArt v 1-specific IgG1 antibodies were also observed in animals vaccinated with rFlaA + rArt v 1 but not in the other treatment groups (Fig. 5B). rArt v 1-specific IgG2a production was only detectable in animals vaccinated with either rFlaA:Artv1 or rFlaA:Artv1 hyp (Fig. 5B). Compared to this prominent induction of IgG antibodies, overall levels of rArt v 1-specific IgE antibodies were very low (OD range: 0.05-0.3, Fig. 5B). Accordingly, Art v 1-specific IgG2a/IgE ratios were only enhanced in animals vaccinated with both fusion proteins (rFlaA:Artv1: two out of six mice, rFlaA:Artv1 hyp : all six mice, Fig. 5B).
While there were no differences in rFlaA-specific IgG1 levels between the groups that received rFlaA (either alone, in a mixture with rArt v 1, or as part of a fusion protein), application of rFlaA:Artv1 resulted in significantly lower levels of rFlaA-specific IgG2a production compared to all other groups (rFlaA, rFlaA + rArt v1, or rFlaA:Artv1 hyp , Fig. 6B). Interestingly, rFlaA sensitization induced production of anti-FlaA IgE antibodies (Fig. 6B). Here, no significant differences in rFlaA-specific IgE levels were detected between animals that had received rFlaA, rFlaA + rArt v 1, or rFlaA:Artv1, while animals that had received rFlaA:Artv1 hyp , did not produce rFlaA-specific IgE antibodies (Fig. 6B). Animals vaccinated with either rFlaA:Artv1 or rFlaA:Artv1 hyp developed significantly higher rArt v 1-specific IgG1 levels than the respective control groups (Fig. 6B). Interestingly, significantly increased levels of rArt v 1-specific IgG2a antibodies were only detected in animals that received rFlaA:Artv1 hyp , but not rFlaA:Artv1 (Fig. 6B). None of the proteins was capable to induce rArt v 1-specific IgE antibodies (Fig. 6B). Finally, our data suggested that the application of both fusion proteins resulted in the predominant induction of TH1-biased immune responses.
In splenocyte cultures re-stimulation with rFlaA resulted in the production of both TH1-(IFN-γ) and TH2-(IL-5 and IL-13) cytokines as well as IL-10 production (Fig. 7B). Interestingly, re-stimulation with rFlaA of splenocytes from animals injected with rFlaA:Artv1 hyp resulted in the strongest reduction of IL-5 and IL-13 levels (Fig. 7B). In animals injected with rFlaA:Artv1 hyp re-stimulation with rFlaA also produced lower amounts of IL-10 and IFN-γ, while cytokine production from splenocytes of animals injected with rFlaA:Artv1 were not different to those observed from the splenocytes of animals that had received rFlaA or rFlaA + rArt v 1 (Fig. 7B). Here, no cytokine secretion was detected upon re-stimulation of splenocyte cultures with rArt v 1 (Fig. 7B).
Both rFlaA:Artv1-and rFlaA:Artv1 hyp -induced IL-10 secretion from mDCs is mediated by mTOR1 activation. Next we addressed the mechanisms by which both fusion proteins modulate immune responses. Since previous studies 20,29 suggested, that the immune modulating properties of flagellin fusion proteins strongly depend on mDC-derived IL-10 secretion we focused on the mechanism of IL-10 induction by both fusion proteins in vitro (Fig. 8).
We observed, that mDC cultures stimulated with rFlaA:Artv1 and rFlaA:Artv1 hyp for up to 72 h revealed a pronounced discoloration of the culture media, which was less pronounced in cells stimulated with the equimolar mixture of both proteins (Fig. 8A). This result suggested a strong metabolic activity of mDCs stimulated with both fusion proteins. This marked increase in overall cellular metabolism resulting from a predominant production of energy by a high rate of glycolysis and lactic acid fermentation is known as the Warburg effect 31 . When quantifying the Warburg effect in mDC cultures stimulated with rFlaA:Artv1 or rFlaA:Artv1 hyp compared to the non-fused mixture of rFlaA + rArt v 1 we observed, that both fusion proteins induced a significantly stronger Warburg effect than the equimolar mixture of both single proteins (Fig. 8A). In line with these results, also glucose consumption from the culture medium and therefore cell metabolic rates were strongly enhanced in mDCs stimulated with both fusion proteins but not the equimolar mixture of rFlaA + rArt v 1 (Fig. 8B,C). Taken together, these results suggested, that stimulation with both fusion proteins resulted in an activation of mDC metabolism. The mTOR1 complex is well described as a master regulator of both cellualr growth and immune responses. Previous results suggested, that flagellin fusion proteins may modulate mDC activation by inducing mTOR signalling (Schülke et al., 2017 JACI).
To verify that mTOR1 is involved in the increased metabolic activity we performed experiments in which we inhibited the mTOR1-complex by pre-treatment of the mDCs with rapamycin, a specific inhibitor of mTOR-activation (Fig. 8). In line with our hypothesis, mTOR-inhibition by rapamycin dose-dependently and significantly suppressed the induced Warburg effect (Fig. 8A), cell metabolic rates (Fig. 8B), glucose consumption from the culture medium (Fig. 8C), as well as IL-10 secretion induced by both fusion proteins (Fig. 8D), while pro-inflammatory IL-6 and TNF-α secretion were not affected by mTOR-inhibition (Fig. 8D). Of note, the Warburg effect and the IL-10 secretion induced by stimulation with rFlaA:Artv1 hyp was slightly (but not significantly) lower than observed for rFlaA:Artv1 (Fig. 8A,D). In line with the theory, that mTOR-induced IL-10 secretion suppresses IL-12p70 production, mDCs stimulated with rFlaA:Artv1 hyp produced significantly higher amounts of IL-12p70 compared to mDCs stimulated with rFlaA:Artv1 (Fig. 8D).
Taken together, these results show that rFlaA:Artv1 and rFlaA:Artv1 hyp modulate immune responses by an mTOR-dependent activation of mDC metabolism, resulting in immunosuppressive IL-10 secretion.
Finally, in an initial set of experiments, we determined the ability of the different constructs to activate human antigen presenting cells. For this, monocyte-derived DCs (moDC) were stimulated with both fusion proteins and the respective controls and checked for cytokine secretion by ELISA (Fig. 9). In accordance with the results obtained for mouse mDC we observed, that rFlaA:Artv1 and rFlaA:Artv1 hyp induced significantly higher levels of IL-10 secretion compared to rFlaA and Art v 1 alone or as a mixture (Fig. 9). Here, overall IL-10 levels induced by rFlaA:Artv1 and rFlaA:Artv1 hyp were comparable. In contrast, IL-6 secretion induced by both fusion proteins was SCIEntIfIC RePoRtS | 7: 11782 | DOI:10.1038/s41598-017-11972-w not statistically different from levels observed for the respective controls. In human moDC IL-12p70 secretion was dose-dependently induced by stimulation with rFlaA, with levels being comparable in moDC stimulated with the mixture of rFlaA + Art v 1 (Fig. 9). Here, both rFlaA:Artv1 and rFlaA:Artv1 hyp induced only slightly higher levels of IL-12p70 secretion, which were only for rFlaA:Artv1-stimulated moDCs statistically different from levels observed after stimulating with the mixture of rFlaA and Art v 1 (Fig. 9).
In summary, both rFlaA:Artv1 and rFlaA:Artv1 hyp were able to induce a prominent secretion of the anti-inflammatory cytokine IL-10 from human dendritic cells, suggesting that these proteins might have the same immune modulating capacity on human cells that we observed in the mouse system.

Discussion
Characterization of FlaA fusion proteins and control proteins. Fusion proteins containing the TLR5ligand flagellin are currently tested in clinical trials for the prevention of influenza infections 32,33 . Furthermore, such fusion proteins are considered as potential vaccine candidates for infectious diseases 24-27 , caries 28,34 , and allergies 20,29 .
In order to further optimize such vaccines candidates for the intervention of allergic diseases, we generated highly pure flagellin fusion proteins incorporating the major mugwort allergen Art v 1, both in its unmodified form as well as hypoallergenic derivative. To generate this hypoallergenic variant we substituted all cysteine residues involved in intramolecular disulfide bonds except Cys 26 in order to maintain the previously identified immunodominant human TC epitope (KCIEWEKAQHGA at position 25 to 36) 10 . The mutation of cysteine residues disrupts intramolecular disulfide bonds and strongly reduces IgE reactivity while preserving T cell reactivity 30 . Consequently, both the hypoallergenic Art v 1 and rFlaA:Artv1 hyp did not show IgE reactivity upon incubation with sera from mugwort allergic patients.
Upon conjugation of the allergen to flagellin, flagellin dominated the CD spectra by its typical alpha-helical secondary structure elements. Here, the secondary structure of rArt v 1 and rArt v 1 hyp showed typical curve progressions for unfolded random coil structures. These findings are in accordance with the available literature showing that the hydroxyproline-rich part of Art v 1 is responsible for the generation of these random coil signals 3 . Analysis of protein aggregation showed, that rFlaA:Artv1 seemed to form several distinct high molecular aggregates, while rFlaA:Artv1 hyp displayed an uniform hydrodynamic radius corresponding to the expected size for a monomeric fusion protein. These differences in aggregation status between rFlaA:Artv1 and rFlaA:Art-v1 hyp may be explained by differences in folding between both fusion proteins, likely caused by both the partial defolding of the fused allergen and the reduced capacity to form intra-and intermolecular disulfide bonds in the Innate immune stimulatory capacity and mechanism of mDC activation. Despite differences in allergen sequence and aggregation status, the hypoallergenic rFlaA:Artv1 hyp retained its capacity to activate the target receptor TLR5 on TLR5-transfected reporter cells as well as the ability to activate both mouse (BALB/cand C57Bl/6-derived) and human dendritic cells resulting in the secretion of both pro-and anti-inflammatory cytokines. Of note, both fusion proteins were able to induce high levels of the anti-inflammatory cytokine IL-10 from both murine and human dendritic cells. These results are in line with our own previous studies showing that a fusion protein consisting of rFlaA and the model allergen ovalbumin was able to induce a strong IL-10 secretion from murine dendritic cells 20,29 , preventing Ova-induced intestinal allergy 29 .
Mechanistically we were able to show the IL-10 secretion, but not the pro-inflammatory cytokine secretion, induced by both fusion proteins, to be inhibited by pre-treatment of the mDCs with rapamycin and therefore to be dependent on mTOR-activation. These results are in accordance with our recently published study where we describe a fusion protein of rFlaA and the major birch pollen allergen Bet v 1 (rFlaA:Betv1) to induce the activation of mammalian target of rapamycin (mTOR) which increased metabolic activity of the stimulated mDCs 35 . Here, as observed for rFlaA:Artv1 and rFlaA:Artv1 hyp , rFlaA:Betv1-mediated IL-10 secretion, but not pro-inflammatory cytokine secretion, was inhibited by rapamycin in mDCs and therefore dependent on mTOR-activation 35 .
Flagellin itself has been described in the literature as effective mucosal adjuvant by enhancing the levels of B7-2 expression in vivo 36 , priming both NF-kappa B-mediated and pro-inflammatory-gene expression in intestinal epithelial cells via TLR5 37 , while also triggering the activation of the NLRC4 inflammasome and subsequent pro-inflammatory cytokine production 38,39 . Here, the promotion of adaptive immunity by flagellin can be effectively driven by either TLR5-mediated activation of NF-κB signaling or NLRC4-mediated inflammasome activation 40 . Besides epithelial cells 37 , CD11c(high)CD11b(high) lamina propria DCs in the small intestine were described as the main TLR5-expressing cells, secreting pro-inflammatory cytokines and inducing the differentiation of naive B cells 41,42 . More recently, flagellin-stimulation was also reported to result in rapid TLR5-dependent mTOR phosphorylation 43 . In this context we speculate, that the immunogenic properties of flagellin are retained in FlaA-containing fusion proteins.
Suppression of TH2 immune responses. Interestingly, while both fusion proteins were able to suppress Art v 1-induced TH2 cytokine secretion from both ex vivo isolated TH2-biased CD4 T cells as well as in the in vivo vaccination model, only rFlaA:Artv1 containing the unmodified allergen sequence was able to efficiently induce IFN-γ secretion in mDC:TC co-cultures. In contrast, rFlaA:Artv1 hyp resulted in a prominent secretion of IL-10 not detected upon stimulation of co-cultures with rFlaA:Artv1. We hypothesize that the observed differences in T cell activation could be explained by either (I) differences in the allergen sequence between rFlaA:Artv1 and rFlaA:Artv1 hyp , (II) differences in protein folding resulting from the disruption of intramolecular cysteine bonds introduced by our mutations, and/or (III) differences in protein uptake and processing due to the aforementioned differences in aggregation status between rFlaA:Artv1 and rFlaA:Artv1 hyp .
While our own studies have shown that the (human) immunodominant T cell epitope of Art v 1 at position 25 to 36 is at least one of the epitopes relevant in mice (data not shown), we cannot exclude the disruption of other mouse T cell epitopes within the modified Art v 1 sequence of the rFlaA:Artv1 hyp fusion protein. However, we observed a robust induction of rArt v 1-specific IgG antibodies upon vaccination with rFlaA:Artv1 hyp . Here, levels of rArt v 1-specific IgG antibodies were not different between rFlaA:Artv1 and rFlaA:Artv1 hyp vaccinated animals, showing that sufficient T cell epitopes were retained within the rFlaA:Artv1 hyp molecule to allow for efficient T cell-dependent B cell activation. The present results provide further evidence that IL-10 secretion induced upon stimulation with flagellin fusion proteins is derived from dendritic cells and only to a very low extend from T cells 29 . Therefore, in vivo application of the fusion proteins was shown to result in suppression of TH2 responses directed against the fused allergen likely mediated by IL-10 production from antigen presenting cells early during vaccination (at the time of initial T cell activation by the fusion protein), but not for example from IL-10 producing regulatory T cells (Schülke, unpublished observation). These results are in line with our observation that IL-10 secretion induced upon in vivo application of a flagellin fusion proteins containing Bet v 1 is produced very rapidly with a maximum after 8 hours post injection and subsequent rapid decline in serum IL-10 levels 35 .

Modulation of humoral immune responses.
While, we did not observe Art v 1-specific antibody responses in vivo upon either vaccination with rArt v 1, the mixture of rFlaA + rArt v 1, or sensitization with rArt v 1 + Alum, analysis of antibody responses in animals vaccinated with rFlaA:Artv1 or rFlaA:Artv1 hyp showed that the activation of the innate immune system also resulted in the efficient induction of adaptive immune responses in vivo. In our experimental setting, at least 4 to 6 injections of either rArt v 1 + Alum or rArt v 1 + Adjuphos are required, in order to induce rArt v 1-specific antibody production (Kuttich, unpublished observation). Therefore, two i.p. injections of rArt v 1 + Alum are likely not sufficient to induce a substantial rArt v 1-specific antibody production but sufficient to induce rArt v 1-specific TH2 responses. In contrast, the induction of rArt v 1-specific IgG1 and IgG2a antibodies was only detected in sera of animals vaccinated with rFlaA:Artv1 and rFlaA:Artv1 hyp , but not the respective controls, suggesting the induction of allergen-specific TH1-biased immune responses in vivo. Additionally, the induction of rFlaA-specific antibodies was retained for both fusion proteins. In the context of allergy treatment, the observed induction of allergen-specific IgG2a antibodies is of particular interest, since these antibodies likely act as blocking antibodies preventing allergic reactions 44 .
In summary, we showed that flagellin:allergen fusion proteins incorporating the major mugwort allergen Art v 1, both in its wild type form and as a hypoallergenic variant, are potent vaccine candidates efficiently suppressing allergen-specific TH2 cytokine secretion both in vitro and in vivo while inducing an mTOR-dependent IL-10 secretion from mDCs. Interestingly, incorporation of the hypoallergenic Art v 1 variant into the flagellin fusion protein, did not interfere with its innate immune stimulatory potential, retaining its capacity to induce secretion of high levels of mTOR-dependent, anti-inflammatory IL-10 from murine and human dendritic cells. Therefore, flagellin:allergen fusion proteins incorporating hypoallergenic allergen variants are promising tools to further improve efficacy and safety of these innovative vaccine candidates for the treatment of allergies.

Material and Methods
Generation of Proteins. For recombinant generation of the major mugwort (Artemisia vulgaris) pollen allergen Art v 1.0101 sequence information was obtained from GenBank (Acc. No: AF493943). Cloning of the fusion proteins was performed using synthetic genes (GeneArt, Regensburg, Germany) encoding for the rFlaA sequence (Acc. No: ×65624) at the N-terminal part and Art v 1/Art v 1 hyp sequences at the C-terminal part, respectively. For the generation of the hypoallergenic fusion protein rFlaA:Artv1 hyp and its respective control Art v 1 hyp , 7 out of 8 total cysteines were exchanged to serines in order to destroy the described disulfide bonds contributing to conformational IgE-epitopes of the allergen 30   performed according to Schülke et al. 45 . Proteins were purified from bacterial extracts by either a combination of immobilized metal ion affinity chromatography and subsequent size exclusion chromatography (rFlaA:Artv1 and rFlaA:Artv1 hyp ) or filtration via a PES-membrane using 50 ml Falcon filtration system (50 kDa cut-off, Sartorius, Göttingen, Germany) followed by cation exchange chromatography (Art v 1 and Art v 1 hyp ) on an ÄKTA FPLC system (GE Healthcare, Munich, Germany). All proteins were dialyzed against PBS (pH = 7.4) for further experiments. Recombinant flagellin A from Listeria monocytogenes (rFlaA, GenBank Acc. No: × 65624) was generated according to 45 . For all proteins endotoxin content was depleted after final protein purification using size exclusion chromatography in combination with either the EndoTrap ® blue system (Hyglos, Bernried, Germany) (rFlaA, rArt v 1, rFlaA:Artv1, rFlaA:Artv1 hyp ) or as described by Bordier et al. 46 (rArt v 1, rArt v 1 hyp ). Protein concentrations were determined using the BCA Protein Assay Kit (Thermo Scientific, Dreieich, Germany). Total yields for final protein preparations were 6.5 mg rArt v 1, 3 mg rArt v 1 hyp , 26 mg rFlaA, 13 mg rFlaA:Artv1, and 12 mg rFlaA:Artv1 hyp . All proteins displayed a purity greater than 98%. SDS-PAGE. SDS-PAGE was performed according to the method described by Laemmli 47 (cross linker C = 5%, total bis/acrylamid 15%) under reducing conditions.

Confirmation of sequence identity by LC-MS E .
Recombinant rFlaA:Artv1 fusion protein was separated by SDS-PAGE and stained with Coomassie Brilliant Blue. Visible protein bands were excised, in-gel digested using trypsin and analyzed by mass spectrometry (nano-UPL nanoESI-MS E as published previously 48 . Differing to this study, MS E data were searched against a UniProt database consisting of reviewed entries of eucaryotic organisms (taxon identifier 2759) and the amino acid sequence of rFlaA:Artv1 fusion proteins.
Patients and sera. Sera of mugwort pollen-allergic individuals were selected based on typical case history, i.e. periodic rhinitis/conjunctivitis during late summer and allergen-specific IgE (CAP-FEIA ≥3) to mugwort pollen (w6; Phadia, Uppsala, Sweden). The study and all used experimental protocols were approved by the ethics committee of the Medical University of Vienna (EK No. 497/2005), all methods were carried out in accordance with the relevant guidelines and regulations. Informed consent was obtained from all subjects.
Western Blot. For IgE immunoblot analysis, proteins were separated by SDS-PAGE and blotted onto nitrocellulose Protran ® membranes (Whatman, Dassel, Germany) according to Towbin et al. 49 . Sera from mugwort allergic patients and non-allergic controls were diluted 1:6 or 1:10 in 100 mM Tris-HCl (pH = 7.4), 100 mM NaCl, 2.5 mM MgCl 2 , 0.05% (v/v) Tween20, 0.1% (w/v) BSA. Bound IgE was detected using a mouse-anti-human IgE-alkaline phospathase (AP) conjugated antibody (BD Pharmingen, Heidelberg, Germany) and visualized with AP Conjugate Substrate Kit (Bio-Rad, Munich, Germany). Circular dichroism spectroscopy. Purified proteins were adjusted to a concentration of 200 µg/ml and dialyzed against 10 mM phosphate buffer. Circular dichroism (CD) spectra were recorded using a JASCO J-810 spectropolarimeter (Jasco, Gross-Umstadt, Germany) with constant N 2 flushing at 20 °C. Measurements were performed in a quartz glass cuvette (1 mm, Hellma, Müllheim, Germany) with a step width of 1 nm and a band width of 1 nm. The spectral range was 180-260 nm at 50 nm/min. Ten scans were accumulated, spectra obtained with buffer were subtracted. The results were expressed as mean residue molar ellipticity [H] MRD and served as indication for formation of secondary structure elements.
Mice. BALB/c mice (Charles River Laboratories, Sulzfeld, Germany,) were kept at the animal facility of the Paul-Ehrlich-Institut under specific pathogen-free conditions. All animal experiments were performed in compliance with the German animal protection law (granting authority: RP Darmstadt, Germany, Approval number: F107/131). France) for 90 minutes and subsequently stimulated with equimolar amounts of rFlaA + rArt v 1, rFlaA:Artv1, or rFlaA:Artv1 hyp for 72 h. Glucose concentrations in culture supernatants were determined 72 h post-stimulation using the Glucose (GO) Assay Kit (Sigma-Aldrich). The metabolic rate was derived mathematically from the measured glucose concentrations by calculating the glucose consumption in % of medium without mDCs (glucose conc. in RPMI1640 = 2 mg/ml).
Statistical analysis. Statistical analysis was performed with GraphPad Prism 5-7 for Mac or Windows using 2-way ANOVA tests with confidence intervals adjusted for multiple comparisons according to Bonferroni. For statistically significant results the following convention was used: *p-value < 0.05, **p-value < 0.01, ***p-value < 0.001.