The Flagellin:Allergen Fusion Protein rFlaA:Betv1 Induces a MyD88− and MAPK-Dependent Activation of Glucose Metabolism in Macrophages

TLR5 ligand flagellin-containing fusion proteins are potential vaccine candidates for many diseases. A recombinant fusion protein of flagellin A and the major birch pollen allergen Bet v 1 (rFlaA:Betv1) modulates immune responses in vitro and in vivo. We studied the effects of rFlaA:Betv1 on bone marrow-derived macrophages (BMDMs). BMDMs differentiated from BALB/c, C57BL/6, TLR5−/−, or MyD88−/− mice were pre-treated with inhibitors, stimulated with rFlaA:Betv1 or respective controls, and analyzed for activation, cytokine secretion, metabolic state, RNA transcriptome, and modulation of allergen-specific Th2 responses. Stimulation of BMDMs with rFlaA:Betv1 resulted in MyD88-dependent production of IL-1β, IL-6, TNF-α, IL-10, CD69 upregulation, and a pronounced shift towards glycolysis paralleled by activation of MAPK, NFκB, and mTOR signaling. Inhibition of either mTOR (rapamycin) or SAP/JNK-MAPK signaling (SP600125) resulted in dose-dependent metabolic suppression. In BMDM and T cell co-cultures, rFlaA:Betv1 stimulation suppressed rBet v 1-induced IL-5 and IL-13 secretion while inducing IFN-γ production. mRNA-Seq analyses showed HIF-1a, JAK, STAT, phagosome, NLR, NFκB, TNF, TLR, and chemokine signaling to participate in the interplay of cell activation, glycolysis, and immune response. rFlaA:Betv1 strongly activated BMDMs, resulting in MyD88−, MAPK−, and mTOR-dependent enhancement of glucose metabolism. Our results suggest macrophages are important target cells to consider during restauration of allergen tolerance during AIT.


Introduction
In the future, novel adjuvants and vaccines may be valuable tools to improve the efficacy, safety, and convenience of allergen immunotherapy. However, for their safe and efficient application, basic knowledge describing their immune modulating capacity, activated cell types, as well as a characterization of the underlying immunological mechanisms is needed. 2

of 27
The Toll-like receptor 5 (TLR5) ligand flagellin is a bacterial motility protein that forms the main body of the bacterial flagellum [1]. Because of its intrinsic immune-activating potential, flagellin was demonstrated to be an effective mucosal adjuvant mediating protective immune responses [2][3][4].
Since flagellin is the only proteinaceous TLR ligand, its application for the generation of recombinant flagellin:antigen fusion proteins is of special interest in vaccine development. Flagellin-containing fusion proteins have already been shown to be both safe and welltolerated in clinical trials [5,6]. Consequently, fusion proteins combining flagellin with different antigens have been investigated for their potential to generate immune protection against different diseases including, among others: influenza [7][8][9], poxvirus [10], West Nile virus [11], and C. tetani [12], and Pseudomonas infections [13].
Moreover, flagellin:antigen fusion proteins are suggested as treatment options for IgE-mediated type I allergies [14][15][16][17]. Kitzmüller and colleagues recently described the enhanced immunogenicity, reduced allergenicity, and intrinsic adjuvant activity of flagellin C:Betv1 fusion proteins containing the major birch pollen allergen Bet v 1 in human monocyte-derived dendritic cells (DCs) and T cells from allergic patients [14].
In theory, adjuvant:antigen fusion proteins have the potential to both boost and modulate antigen-specific immune responses by targeting the fused antigen to immune cells in vivo that express receptors for the respective adjuvant. In addition, adjuvantmediated immune cell activation influences the processing and increases the presentation of antigen-derived peptides in the context of the adjuvant-mediated immune cell activation [4]. Concordantly, our work has demonstrated that the fusion of allergens to the TLR5-ligand flagellin was able to suppress allergen-specific Th2 responses both in vitro and in vivo via the induction of both IL-10-mediated anti-inflammatory responses and Th1 responses [15][16][17][18]. However, the induction of overreaching Th1 responses may also have detrimental effects [19]. Therefore, the balanced induction of both pro-and antiinflammatory responses by rFlaA:Betv1 may have advantages compared to exclusively triggering Th1-biased responses.
In our previous work, the fusion protein rFlaA:Betv1, consisting of flagellin A from Listeria monocytogenes and the major birch pollen allergen Bet v 1, was shown to suppress allergic sensitization in vivo [15]. These immune-modulatory effects were paralleled by a pronounced rFlaA:Betv1-mediated activation of myeloid dendritic cells (mDCs) in vitro, secretion of both pro-(IL-6, IL-12, and TNF-α) and anti-inflammatory (IL-10) cytokines, and a mammalian target of rapamycin (mTOR)-dependent metabolic switch towards glycolysis [15]. In this context, the contribution of both immune cell metabolism to the overall immune responses induced by such fusion proteins and other immune cell types needs further investigation.
Macrophages are both the main effector cells of the innate immune system and professional antigen-presenting cells (together with DCs and B cells). Therefore, macrophages control the induction and regulation of T cell immune responses via the uptake, processing, and presentation of antigens to antigen-specific T cells [20,21]. Epigenetics, cell survival, and the tissue microenvironment are the three main pathways controlling the phenotypical polarization of macrophages [22].
M2a macrophage differentiation is induced by the Th2 cytokines IL-4 and IL-13; IL-1 receptor ligands and immune complexes were reported to induce M2b macrophages. IL-10, TGFβ, and glucocorticoids were reported to induce M2c macrophages, and finally, TLR antagonists were reported to induce M2d macrophage differentiation [28]. Here, M2a Cells 2021, 10, 2614 3 of 27 macrophages secrete high levels of the Th2 cytokine IL-13 as well as other chemokines, which promote Th2 responses and induce eosinophil infiltration [28]. Both M2b and M2c macrophages are capable of secreting anti-inflammatory IL-10 in order to regulate immune responses. The M2d phenotype is known as a tumor-associated macrophage that regulates tumor metastasis [26]. While M1 and M2 macrophages are closely related to Th1 and Th2 responses, several studies have indicated the role of macrophages in allergic responses, especially in asthma [28][29][30][31][32]. In general, the M1 phenotype is the major effector macrophage in non-allergic asthma [31], and patients with higher frequencies of M1 cells have less severe clinical allergic reactions [29]. In contrast, M2 cells, especially of the M2a phenotype, are suggested to be the major effector macrophages in allergic asthma because they can promote Th2 responses, promote the secretion of histamine by interaction with antigenspecific Th2 cells, and induce eosinophil migration [28][29][30][31][32]. Therefore, understanding the effects of rFlaA:Betv1 on macrophages could further improve our knowledge of the immune-modulating properties of this promising vaccine candidate for type I allergies.

Determination of Beta-Glucans
To exclude the immunological effects of TLR2 ligands in the used preparation of rFlaA:Betv1, the concentration of (1→3) beta-glucans contained within the used rFlaA:Betv1 preparation was determined using the Endosafe-PTS Portable Test System (Charles River Laboratories, Cologne, Germany) according to the manufacturer's recommendations.

Mice
BALB/c, C57BL/6 mice, TLR5 −/− , and MyD88 −/− mice (C57BL/6 background, all from Jackson Laboratories, Bar Harbor, Maine, ME, USA) were bred at the animal facility of the Paul Ehrlich Institute 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).

In Vitro Generation of Mouse Bone Marrow-Derived Macrophages, Stimulation, and Flow Cytometry
Mouse BMDMs were generated from bone marrow by culture in RPMI 1640 medium (Gibco, Karlsruhe, Germany) supplemented with 10% FCS, 1 mM sodium pyruvate, 10 mM HEPES, 100 U/mL penicillin, 100 µg/mL streptomycin, 1 mM L-glutamine, 0.1 mM 2mercaptoethanol, and 20 ng/mL of mouse recombinant macrophage colony-stimulating factor (rmM-CSF, PeproTech, Hamburg, Germany). For differentiation into BMDMs, 2 × 10 6 cells were seeded per 100 mm dish in 10 mL complete medium. On day 3, 10 mL of complete medium were added per culture dish. On days 6 and 8, 10 mL medium were exchanged and on day 10 M-CSF concentration was reduced to 12.5 ng/mL.

Analysis of Cell Metabolic State
The Warburg effect, glucose consumption, and metabolic rate in stimulated BMDM cultures were determined according to [15].

Metabolic Flux Analysis
For metabolic flux analysis, 1 × 10 5 BMDMs per well were seeded in Seahorse XF96 cell culture microplates (V3-PS, TC-treated, Agilent, Santa Clara, CA, USA). The next day, the medium was exchanged, and cells were stimulated as indicated using the different proteins. Seahorse XF Real-Time ATP rate assays, Seahorse XF Glycolysis, and Seahorse XF Cell Mito Stress Tests were performed according to the manufacturer's recommendations (Agilent). Cycle numbers were as follows: 4 cycles of baseline measurement, 14 cycles of stimulation with the different proteins, and 8 cycles with either oligomycin, rotenone/ antimycin A (Rot/AA), and 2-deoxy-glucose (2-DG), respectively (1 cycle = 3 min mixing plus 3 min measuring). Post-measurement, samples were normalized to total protein content via BCA (Thermo Fischer Scientific) and analyzed using Wave Desktop software (Agilent) and Graphpad Prism v8. Glycolytic Stress and Mito Stress results were analyzed using the respective report generator sheets according to the manufacturer's recommendations (Agilent).

RNA-Seq and Bioinformatics
For RNA-Seq analyses, 3.2 × 10 5 BMDMs/mL were stimulated in 24-well plates with either LPS as a positive control or the indicated equimolar concentrations of rFlaA + rBet v 1 or rFlaA:Betv1 for 48 h. Subsequently, BMDMs were harvested (4 biological replicates per condition) and frozen in Trizol (Invitrogen, Thermo Fisher Scientific) before total RNA isolation with the Direct-zol RNA Miniprep Kit (Zymo Research, Freiburg, Germany). We harnessed a modified version of the NNSR priming method [35] to prepare stranded, Illumina-compatible libraries as follows: we used the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB) to isolate mRNA from 1 µg total RNA per sample. cDNA was synthesized with Superscript IV in the presence of 4 µg Actinomycin D and RiboLock RNase inhibitor (Thermo Fisher Scientific) with the NNSR_RT primer (see Table 1   The raw sequencing reads were quality-, length-, and adapter-trimmed with fastp [36]. We used STAR [37] to align the trimmed reads to the human genome (hg38 assembly) and to count the mapped reads on gene level. The quantitative evaluation of the counts, differential gene expression, and the accompanying statistical analyses were performed in the R environment (https://www.r-project.org, version 3.5.0. (accessed on 23 April 2018) using the DESeq2 package (version 1.28.3)) [38]. KEGG pathway analysis [39] and gene set enrichment analysis were performed with the clusterProfiler package [40]. Data were deposited in the SRA database (accession number: PRJNA755840).

Statistical Analysis
Statistical analyses were performed with GraphPad Prism v6 to v8 for either Mac or Windows using 2-way ANOVA tests with confidence intervals adjusted for multiple comparisons according to either Bonferroni or Tukey. For statistically significant results, the following convention was used: * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.
To exclude potential immune-activating effects of the residual amounts of LPS contained within the used rFlaA:Betv1 preparation, we established thresholds for LPS-induced BMDM activation (Repository Figure S3A). Here, the threshold for an LPS-induced Warburg effect was between 1 and 10 ng. Increased glucose consumption and significant secretion of IL-10 were observed starting at 10 ng; secretion of IL-6 started at 1 ng, while IL-1β-and TNF-α-secretion were detected after stimulation with 1 µg of LPS (Repository Figure S3B,C). In addition, stimulation of BMDMs with the amounts of LPS contained within the used concentrations of rFlaA:Betv1 (0.59 pg LPS contained within 0.247 µg/mL rFlaA:Betv1, 5.9 pg LPS in 2.47 µg/mL rFlaA:Betv1, and 59 pg LPS in 24.7 µg/mL rFlaA:Betv1) neither resulted in activation of BMDM metabolism nor cytokine secretion (Repository Figure S3B).
To further exclude effects of LPS or potentially contaminating TLR2 ligands in the used rFlaA:Betv1 preparation, we used HEK293 cell lines stably transfected with either mouse TLR2, TLR4, or TLR5 as well as non-TLR-transfected HEK293 control cells (Repository Figure S4A). Here, rFlaA:Betv1 exclusively induced TLR5 activation, as indicated by a dose-dependent release of IL-8 from these cells (Repository Figure S4B). To further exclude the presence of TLR2 ligands in the rFlaA:Betv1 preparation, levels of (1→3) beta-glucans were determined. Detection of (1→3) beta-glucans in the used rFlaA:Betv1 preparation showed the levels of (1→3) beta-glucans to be below 0.8 pg/µg of protein, which is safely below the low ng amounts of different TLR2-ligands (LPS, mycoplasma diacylated lipoprotein, lipomannan, lipoarabinomannan, and Pam 3 CysK 4 ) that were previously reported to activate TLR2 on BMDMs [41]. Therefore, we could show that neither the residual LPS amounts nor TLR2 ligands contribute to the rFlaA:Betv1-mediated activation of BMDMs.
In summary, stimulation of BMDMs with rFlaA:Betv1, and, to a lower degree, with rFlaA and rFlaA + rBet v 1, resulted in a pronounced activation of both BMDM metabolism and pro-and anti-inflammatory cytokine secretion. Figure 1. Activation of BMDM metabolism by rFlaA:Betv1 is accompanied by both pro-and anti-inflammatory cytokine secretion. C57BL/6 BMDMs were differentiated from mouse bone marrow and stimulated with the indicated equimolar protein amounts or LPS as a positive control for 96 h (A). Supernatants were analyzed for the induced Warburg effect and glucose consumption (B) as well as cytokine secretion by ELISA (C). Data are mean results of three independent experiments ± SD with two technical replicates per experiment. Statistical significance indicated as: n.s. p-value > 0.05, * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.

rFlaA:Betv1 Triggers a Pronounced Shift Towards Glycolysis in BMDMs
In-depth analysis of the BMDM metabolic state using Seahorse technology ( Figure 2A) revealed rFlaA, rFlaA + rBet v 1, and rFlaA:Betv1 to dose-dependently increase glycolysis (detected by enhanced extracellular acidification rates (ECAR)), while at the same time reducing mitochondrial respiration (reduced oxygen consumption rates (OCR), Figure 2B). In contrast, rBet v 1 had only minimal effects on BMDM metabolism ( Figure 2B). Further inhibition of the mitochondrial ATP synthase by oligomycin and of electron transport chain complexes by rotenone/antimycin A (Rot/AA) resulted in compensatory increases in ECAR, which were less pronounced in rFlaA:Betv1-stimulated BMDMs compared to the controls ( Figure 2B). Treatment with the competitive inhibitor of glucose-6-phosphate, generation of 2-deoxy-glucose (2-DG) completely suppressed extracellular acidification ( Figure 2B). In line with these results, analysis of the glycolytic state showed that rFlaA:Betv1-stimulated BMDMs have similar glycolytic capacities (capacity to produce energy by glycolysis) but higher glycolytic rates (amount of energy generated by glycolysis per minute and cell) compared to BMDMs stimulated with rFlaA + rBet v 1. In comparison, BMDMs stimulated with equimolar amounts of rFlaA + rBet v 1 have higher glycolytic reserves (capacity to further increase energy production by glycolysis) while no significant differences in non-glycolytic acidification (acidification of the extracellular medium by processes other than glycolysis) were observed between treatment groups ( Figure 2C). In contrast to this, detailed analysis of mitochondrial function via Mito Stress tests revealed no differences in basal respiration, spare respiratory capacity, mitochondrial ATP production, proton leak, and non-mitochondrial respiration between rFlaA + rBet v 1and rFlaA:Betv1-stimulated cells (Repository Figure S5).
Taken together, the metabolic phenotyping of rFlaA:Betv1-stimulated BMDMs suggested the fusion protein induced a pronounced shift towards increased levels of glycolysis while mitochondrial respiration was suppressed. In order to study the signaling mechanisms contributing to rFlaA:Betv1-mediated BMDM activation, BMDMs were differentiated from either C57BL/6, TLR5-, or MyD88deficient mice, stimulated with the fusion protein and the respective controls, and analyzed to assess their metabolic state and cytokine secretion profile ( Figure 3A).
Interestingly, activation of BMDM metabolism (Warburg effect and glucose consumption) by both rFlaA and rFlaA:Betv1 was shown to be partially TLR5-dependent ( Figure 3B). MyD88-deficient BMDM did not show increased metabolic activity after stimulation with either of the tested proteins ( Figure 3B). In line with these results, cytokine secretion induced by either rFlaA, rFlaA + rBet v 1, or rFlaA:Betv1 was shown to be abrogated in MyD88-deficient BMDMs ( Figure 3C). Among the tested cytokines, only rFlaA:Betv1induced IL-10 and TNF-α showed a substantial TLR5 dependency (41% and 87% reduction in TLR5-deficient mice compared to C57BL/6 BMDMs, respectively) while secretion levels of all other cytokines were unaffected by TLR5 deficiency ( Figure 3C).
In summary, using BMDMs differentiated from either TLR5− or MyD88-deficient bone marrow, the activation of glucose metabolism and cytokine secretion was shown to be abrogated in the absence of MyD88 while being partially TLR5 dependent for the secretion of IL-10 and TNF-α. and analyzed for extracellular acidification rates (ECAR) and oxygen consumption rates (OCR) using Seahorse technology (B). 14 cycles (84 min) poststimulation, ATP synthase, the electron transfer chain, and glycolysis were inhibited by sequential injection of oligomycin, rotenone/antimycin A (Rot/AA), and 2-deoxy-glucose (2-DG), for 8 cycles (48 min) each, respectively (B). Glycolytic stress test was performed in BMDMs stimulated with either rFlaA + rBet v 1 or rFlaA:Betv1 (both equimolar to 4 µg of rBet v 1) (C). Levels of glycolysis, glycolytic capacity, glycolytic reserve, and non-glycolytic respiration were analyzed by Seahorse technology using the "XF Glycolysis Stress Test Report Generator" according to the manufacturer's recommendations (C). Data are representative results of three independent experiments (with three to four technical replicates per experiment) that showed similar results. Statistical significance indicated as: n.s. p-value > 0.05, * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.  Stimulation of C57BL/6 BMDMs with rFlaA:Betv1 resulted in higher up-regulation of CD69 compared to either rFlaA− or rFlaA + rBet v 1-stimulation ( Figure 4B and quantified in Figure 4C). Of note, both rFlaA− and rFlaA:Betv1-induced CD69 upregulation was partially, but not significantly, reduced in TLR5-deficient BMDMs, while being completely abrogated in MyD88-deficient BMDMs ( Figure 4B,C). Stimulation with rBet v 1 alone in either tested mouse strain did not result in increased CD69 expression ( Figure 4B,C).
Taken together, we observed rFlaA:Betv1 to induce a stronger activation of MAPK, NFκB, and mTOR signaling in BMDMs compared to either rFlaA alone or the mixture of both proteins. rFlaA:Betv1-induced activation of MAPK and NFκB signaling was largely TLR5− and MyD88-independent, while p70 S6 kinase activation was only abrogated in MyD88-deficient BMDMs.

Activation of BMDM Metabolism and IL-10 Secretion are Partly mTOR-Dependent while Pro-Inflammatory Cytokine Secretion Is mTOR-Independent
To study the contribution of either mTOR and/or different MAP kinases to rFlaA:Betv1mediated activation of BMDM metabolism and cytokine secretion, BMDMs were preincubated with either rapamycin to inhibit mTOR activation ( Figure 5A) or different MAPK inhibitors ( Figure 6A). Toxicity of all tested inhibitors on BMDMs was excluded by live/dead staining (Repository Figure S6). Figure 5. rFlaA:Betv1-induced activation of C57BL/6 BMDM metabolism and IL-10 secretion are partly mTOR-dependent, while pro-inflammatory cytokine secretion is mainly mTOR-independent. C57BL/6 BMDMs were differentiated from mouse bone marrow, pre-treated with the indicated rapamycin (mTOR inhibitor) concentrations for 90 min, and subsequently stimulated with rFlaA:Betv1 for an additional 6, 12, 24, or 96 h (A). Cells were analyzed for their metabolic state and cytokine secretion by ELISA after either 6 to 24 h (B) and 96 h (C). Data are mean results of three independent experiments ± SD, with two technical replicates per experiment. Statistical significance indicated as: n.s. p-value > 0.05, * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.
To inhibit mTOR complex 1 (mTORC1) activation, BMDMs were pre-treated with rapamycin and checked for the Warburg effect, glucose consumption, and cytokine secretion after either 6, 12, 24, or 96 h ( Figure 5B,C). Interestingly, dependent on the time, the engagement of mTOR had different effects on IL-10 secretion.
While up to 24 h post-stimulation a rFlaA:Betv1-induced Warburg effect and increased glucose consumption were not yet detectable, mTOR inhibition dose-dependently suppressed rFlaA:Betv1-induced IL-10 secretion at all investigated time points. Here, proinflammatory cytokine secretion was either unaffected (IL-6 & TNF-α) or even increased (IL-1β, Figure 5B). In contrast to this, at 96 h post-stimulation, inhibition of mTORC1 activation slightly, but not significantly, inhibited the rFlaA:Betv1-induced Warburg effect and reduced glucose consumption from the culture medium ( Figure 5C). With the exception of a dosedependent 2.4-fold increase in rFlaA:Betv1-induced IL-1β secretion, rFlaA:Betv1-induced IL-10, IL-6, and TNF-α secretion were all unaffected by mTOR-inhibition ( Figure 5C). To compensate for a potential degradation of rapamycin in the BMDM cultures, we added 2.5 nM of rapamycin daily for 4 days and compared the results to a one-time application of 10 nM rapamycin (as was done before in Figure 5, Repository Figure S7A). Here, in accordance with our previous results, neither a single application of 10 nM rapamycin at the beginning of the culture nor sequential adding of rapamycin over the whole culture time had any significant effects on either the rFlaA:Betv1-induced Warburg effect, glucose consumption, or cytokine secretion (Repository Figure S7B). Therefore, these data suggest that, while rFlaA:Betv1-induced pro-inflammatory cytokine secretion is largely mTOR-independent, early rFlaA:Betv1-induced IL-10 secretion is mediated via mTOR. At later time points, mTOR-independent factors seem to drive IL-10 secretion.

MAP Kinase Signaling Contributes to rFlaA:Betv1-Induced Activation of BMDM Metabolism as well as Both Pro-and Anti-Inflammatory Cytokine Secretion
To investigate the contribution of different types of MAPK to rFlaA:Betv1-induced activation of BMDM metabolism and cytokine secretion, BMDMs were pre-treated with either the p38 MAPK inhibitor SB202190, the p42/44 MAPK inhibitor U0126, or the SAP/JNK MAPK inhibitor SP600125 ( Figure 6). Interestingly, inhibition of different types of MAP kinases had differential effects on rFlaA:Betv1-induced activation of BMDM metabolism and cytokine secretion ( Figure 6A).
Inhibition of p38 MAPK by pre-treatment with SB202190 and p42/44 MAPK by pretreatment with U0126 had no effect on either the rFlaA:Betv1-induced Warburg effect or glucose consumption ( Figure 6B, left and middle lane). In contrast to this, inhibition of SAP/JNK MAPK by SP600125 dose-dependently and highly significantly suppressed the rFlaA:Betv1-induced Warburg effect and glucose consumption ( Figure 6B, right lane).

rFlaA:Betv1 Suppresses Th2 Responses from Bet v 1-specific CD4 + T Cells In Vitro
To test the immune-modulatory capacity of rFlaA:Betv1-stimulated BMDMs, cells were cultured with CD4 + T cells isolated ex vivo from rBet v 1 plus alum-sensitized BALB/c mice ( Figure 7A). Co-cultures were re-stimulated with either (I) rBet v 1 to induce rBet v 1-specific recall responses, (II) equimolar amounts of rFlaA:Betv1 to check for rFlaA:Betv1induced metabolic changes and cytokine secretion in BMDM:CD4 + T cell co-cultures, (III) rBet v 1 plus the equimolar mixture of rFlaA + rBet v 1 to analyze the effect of the protein mixture on rBet v 1-induced cytokine responses, or (IV) rBet v 1 plus rFlaA:Betv1 to investigate the effect of the fusion protein on rBet v 1-induced cytokine responses. Interestingly, stimulation with rBet v 1 only induced a very slight Warburg effect in BMDM:CD4 + T cell co-cultures, which was stronger after stimulation with the mixture of rFlaA + rBet v 1 (equimolar to 10 µg rBet v 1), while rFlaA:Betv1 induced a strong Warburg effect. even in the lowest stimulation concentration ( Figure 7B). Results of glucose consumption paralleled the results observed for the Warburg effect (with high Warburg effects resulting in high glucose consumption, Figure 7B).

rFlaA:Betv1 Induces a Transcriptional Shift Towards HIF-1a-Mediated Glycolytic Metabolism in BMDMs
To (I) confirm both the activation of glycolytic metabolism and BMDM activation and (II) investigate the regulation of other signaling pathways in rFlaA:Betv1-stimulated BMDMs in an unbiased way, BMDMs were stimulated with either LPS, rFlaA + rBet v 1, or rFlaA:Betv1 for 48 h, and whole-cell RNA expression profiles were generated using MiSeq technology ( Figure 8A). In addition, samples were collected for Western blot analysis 2, 6, 24, and 48 h post-stimulation ( Figure 8A).
Principal component analyses showed the biological replicates for each stimulation condition to closely group with each other, while the different stimulation conditions formed clearly separated groups ( Figure 8B). Here, rFlaA:Betv1-stimulated BMDMs showed similar cell activation capacity as LPS-stimulated cells ( Figure 8B).
Finally, activation of JAK-STAT-HIF-1a signaling, glycolysis, and the downregulation of mitochondrial respiration observed in transcriptomic samples were confirmed by Western blot ( Figure 8D). Here, we observed a phosphorylation of STAT3 2, 6, and 24 h post-stimulation, which was more pronounced in LPS− and rFlaA:Betv1-stimulated BMDMs, and followed by a strong upregulation of HIF-1a starting 24 h post-stimulation after the peak of STAT3 phosphorylation at 6 h ( Figure 8D). We also observed a stronger expression of the glucose transporter Glut-1 in all stimulation conditions 48 h poststimulation and a stronger expression of the glycolysis-promoting enzyme 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3) in LPS− and rFlaA:Betv1-stimulated BMDMs 24 h post-stimulation. In accordance with the observed switch to glycolytic metabolism, expression levels of the mitochondrial enzyme aconitase 2 (ACO2) were lower in LPS− and rFlaA:Betv1-stimulated BMDMs both 24 and 48 h post-stimulation ( Figure 8D). Taken together, the results obtained by RNA-seq analyses confirm a pronounced shift towards anaerobic metabolism in rFlaA:Betv1−, LPS−, and, to a lesser extent, rFlaA + rBet v 1-stimulated BMDMs.
In summary, the presented study reports the immune metabolic effects of a novel type of vaccine candidate generated by combining the TLR5 ligand flagellin with the major birch pollen allergen Bet v 1 (which serves a model antigen). Our results suggest that stimulation of BMDMs with rFlaA:Betv1 triggered a MyD88-dependent, but only partly TLR5-dependent, cytokine production, stronger activation of HIF-1a, MAPK, NF B, and mTOR signaling (with the activation of MAPK being located upstream of mTOR signaling), and increased glucose metabolism. In BMDM and CD4 + T cell co-cultures, rFlaA:Betv1 stimulation significantly suppressed Bet v 1-induced IL-5 and IL-13 secretion while inducing IFN-γ production (summarized in Figure 9). Therefore, macrophages contribute to the strong immune-modulating effects of rFlaA:Betv1 previously observed in vivo [15] and should be considered important target cells contributing to the re-establishment of allergen tolerance in AIT.
Finally, our results suggest that both the fusion of antigens to immune-activating molecules and the resulting structural alterations in the fused antigen can increase its immunogenicity, resulting in the induction of distinct immune responses. Therefore, such adjuvant:antigen fusion proteins likely are valuable tools to induce distinct (in our case balanced Th1) immune responses against antigens with otherwise either low immunogenicity or a high risk of triggering potentially harmful Th2 responses. Figure 9. Suggested molecular signaling events contributing to rFlaA:Betv1-mediated activation of BMDMs. Stimulation of BMDM with aggregated rFlaA:Betv1 results in both TLR5-dependent and -independent uptake, leading to the activation of both MyD88− and MAPK-mediated signaling cascades. rFlaA:Betv1-induced pro-inflammatory IL-6 and TNF-α secretion were shown to be MyD88− and MAPK-dependent, likely promoting the induction of IFN-γ-producing Th1 cells in BMDM:TC co-cultures. Western blot analysis confirmed rFlaA:Betv1 to trigger STAT3 activation, increasing the expression of HIF-1a, which is likely to be responsible for the observed IL-1β secretion. Metabolically, we found a MyD88-de- Figure 9. Suggested molecular signaling events contributing to rFlaA:Betv1-mediated activation of BMDMs. Stimulation of BMDM with aggregated rFlaA:Betv1 results in both TLR5-dependent and -independent uptake, leading to the activation of both MyD88− and MAPK-mediated signaling cascades. rFlaA:Betv1-induced pro-inflammatory IL-6 and TNF-α secretion were shown to be MyD88− and MAPK-dependent, likely promoting the induction of IFN-γ-producing Th1 cells in BMDM:TC co-cultures. Western blot analysis confirmed rFlaA:Betv1 to trigger STAT3 activation, increasing the expression of HIF-1a, which is likely to be responsible for the observed IL-1β secretion. Metabolically, we found a MyD88-dependent phosphorylation of the mTOR target protein p70 S6 kinase, which together with HIF-1a likely mediated the observed shift towards enhanced glycolysis with increased levels of the glucose transporter Glut-1 and PFKFB3, and the induction of the Warburg effect while being paralleled by a decreased expression of mitochondrial proteins such as ACO2 and reduced oxygen consumption rates, suggesting a disrupted Krebs cycle in these cells. Finally, TLR5, MyD88, and MAPK activation were shown to be involved in the rFlaA:Betv1-induced production of the anti-inflammatory cytokine IL-10 that likely caused the observed suppression of rBet v 1-induced Th1 and Th2 responses in BMDM:TC co-cultures.

Discussion
In the present study, we analyzed the activation, metabolic state, and transcriptional state of BMDMs, as well as their immune-modulatory properties upon stimulation with a flagellin:allergen conjugate (rFlaA:Betv1), which was recently shown in preclinical studies to prevent allergic Th2 responses [15]. Here, the allergen Bet v 1 serves as a model antigen that is conjugated to the TLR5 ligand flagellin FlaA, which acts as an adjuvant activating the immune system. Thereby, FlaA may modulate the immunological properties of the conjugated antigen while also targeting the antigen to innate immune cells.
Currently, the effects of flagellin-containing fusion proteins on macrophages, which serve as important innate immune cells, are less described. So far, only one study by Verma et al. has described the insertion of a cytochrome C molecule into the flagella of Salmonella dublin to efficiently trigger processing and presentation of cytochrome-derived peptides by mouse peritoneal macrophages [42]. This prompted us to investigate the effect of FlaA and FlaA-containing fusion proteins on mouse macrophages.

rFlaA:Betv1 Induces a Partly TLR5-Dependent Activation of BMDMs Characterized by Both Pro-and Anti-Inflammatory Cytokine Secretion
We observed a stronger activation of mouse BMDMs by rFlaA:Betv1 compared to the respective controls. However, both stimulation with rFlaA and rFlaA + rBet v 1 resulted in a weaker, but still clearly detectable, activation of cytokine secretion and glycolytic metabolism in BMDMs. In line with the available literature [2][3][4], these results confirmed the adjuvant effect of non-conjugated flagellin.
BMDM activation by rFlaA:Betv1 was characterized by strongly increased pro-(IL-1β, IL-6, TNF-α) and anti-inflammatory (IL-10) cytokine secretion. In terms of the net outcome of this mixed pro-and anti-inflammatory response, the observed suppression of rBet v 1induced Th2 cytokine secretion from BMDM:CD4 + T cell co-cultures suggested rFlaA:Betv1mediated anti-inflammatory signaling (at least in higher stimulation concentrations) to be more pronounced.
Interestingly, we observed a significantly increased secretion of IL-1β upon inhibition of mTOR in rFlaA:Betv1-stimulated BMDMs both 24 and 96 h post-stimulation. Tang et al. reported similar results from tuberous sclerosis complex 1 (TSC-1)-deficient macrophages, which display constitutively active mTOR signaling [46]. Here, impaired (LPS-induced) IL-1β secretion from tuberous sclerosis complex 1 (TSC-1)-deficient macrophages could be increased by either inhibition of mTOR activation or mTOR deletion [46]. Mechanistically, mTOR-mediated downregulation of the CCAAT enhancer-binding protein (C/EBPβ) was shown to critically contribute to the reduction in IL-1β production [46]. These results suggest that, both in the model of Tang et al. and in rFlaA:Betv1-stimulated BMDMs, TSC-1 may promote IL-1β production via the C/EBPβ pathway, while activation of mTOR suppresses both TSC-1 itself and TSC-1-dependent downstream processes (including IL-1β production) [46].
In our previous publication describing the molecular characterization of rFlaA:Betv1, we reported rFlaA:Betv1 to be aggregated [15]. Here, rFlaA:Betv1 formed high molecular aggregates, which were only visible under native PAGE conditions [15]. Increased uptake of these aggregates in myeloid dendritic cells was shown to be TLR5-independent while being essential for the induction of rFlaA:Betv1-mediated IL-10 and IL-6 secretion [15]. Therefore, aggregation was previously shown to be an important factor determining the immunogenicity of the fusion proteins.
Mechanistically, we assume the stronger BMDM-activating potential of the fusion protein compared to both single proteins to result from: (i) the previously observed aggregation of rFlaA:Betv1 [15], which results in higher densities of the fusion protein on the surface of BMDMs, higher protein uptake, and therefore stronger activating signals transmitted to the respective BMDMs nucleus, (ii) changes in BMDM metabolic state due to the strongly activating signal provided by rFlaA:Betv1, and (iii) (for the induction of adaptive immune responses by rFlaA:Betv1-stimulated BMDMs) differences in antigen processing upon fusion to flagellin already reported for rOva contained within a rFlaA:Ova fusion protein [18].
Taken together, these results suggest that both TLR5-dependent and -independent processes contribute to rFlaA:Betv1-mediated BMDM activation, while MyD88 is a key molecule in rFlaA:Betv1-mediated macrophage activation.

rFlaA:Betv1-Stimulated BMDMs Efficiently Modulate Allergen-Specific T Cell Responses
Considering that the activation of BMDMs by rFlaA:Betv1 is antigen unspecific, we checked for the effect of rFlaA:Betv1-stimulated BMDMs on Th2-biased Bet v 1-specific T cells in an ex vivo co-culture model. Here, both rFlaA + rBet v 1 and rFlaA:Betv1 efficiently suppressed rBet v 1-induced Th2-cytokine secretion, while only rFlaA:Betv1 boosted IL-10 and IFN-γ secretion, once again confirming the pronounced immune modulatory capacity and Th1-inducing capacity of the fusion protein. One weakness of the employed co-culture setup is that the cellular source of the secreted cytokines (BMDMs or T cells) cannot be exactly determined. However, despite this limitation, our data provide evidence that simultaneous targeting of adjuvant and antigen to BMDMs efficiently promotes modulation of antigen-specific effector T cell responses. While in this study we investigated rBet v 1-specific T cell responses, our previous work on the model antigen ovalbumin [16,18] and the major mugwort allergen Art v 1 [17] suggests that the findings can also be transferred to other antigens.

rFlaA:Betv1 Activates Both Glycolytic Metabolism and HIF-1a Signaling in BMDMs
A further objective of this study was to explore the mechanisms underlying the observed immune modulation in more detail. Previous reports suggested mTOR to be engaged in rFlaA:Betv1-mediated cytokine responses [15]. rFlaA:Betv1 also triggered changes in the metabolic phenotype of the stimulated BMDMs. In activated immune cells, distinct metabolic changes not only provide the energy needed by these cells to proliferate and fulfill their effector functions, but are also used to generate important immunological active effector molecules (e.g., prostaglandins, ROS, NOS, or itaconate generated from a disrupted Krebs cycle in highly glycolytic cells) [47]. Therefore, cellular metabolism and signaling pathways classically associated with the activation and effector function of immune cells have recently been recognized to be intimately intertwined, regulating each other, and thereby shaping the overall immune responses induced.
In our experimental setting, HIF-1a, which was shown to be essential for glycolytic metabolism in macrophages, seems to be of particular importance. HIF-1a-deficient macrophages lack both lactate and ATP production after LPS stimulation, and impaired glycolysis was shown to profoundly reduce the clearance of bacteria and fungi [56,57]. In contrast to M1 macrophages, activated M2 macrophages were reported to display STAT6− and PGC-1β-dependent decreases in glycolytic activity, while mitochondrial oxygen consumption rates and fatty acid oxidation were increased [58,59]. In line with the suggested M1 phenotype, stimulation of BMDMs with rFlaA:Betv1 resulted in a pronounced activation of HIF-1a signaling, which was not observed in BMDMs stimulated with equimolar amounts of rFlaA + rBet v 1. These results suggest the induction of HIF-1a signaling being critical for rFlaA:Betv1-mediated BMDM activation. Remarkably, succinate accumulating from a disrupted Krebs cycle in LPS-stimulated macrophages was shown to promote both glycolysis and IL-1ß production via the stabilization of HIF-1a (reviewed in [47]), thereby linking the activation of glycolytic metabolism to the activation of HIF-1a signaling. Interestingly, this newly discovered activation of HIF-1a signaling by rFlaA:Betv1 might further explain the previously observed induction of Th1-biased immune responses by flagellin:allergen fusion proteins in vitro and in vivo [15][16][17][18], as the activation of HIF-1a in macrophages was recently described to promote induction of Th1 responses [60]. Besides, M1 macrophages are known to produce pro-inflammatory cytokines, such as IL-12, TNF-α, IFN-γ, and IL-6, and can further induce Th1 responses [61]. The production of these cytokines was also reported to be downregulated in macrophage-specific HIF-1a knockout mice upon β-aminopropionitrile stimulation [62]. Indeed, we observed that rFlaA:Betv1 boosted production of the Th1 cytokine IFN-γ secretion even from BMDM:CD4 + T cell co-cultures with Th2-biased T cells (Figure 7).
In summary, the presented study reports the immune metabolic effects of a novel type of vaccine candidate generated by combining the TLR5 ligand flagellin with the major birch pollen allergen Bet v 1 (which serves a model antigen). Our results suggest that stimulation of BMDMs with rFlaA:Betv1 triggered a MyD88-dependent, but only partly TLR5-dependent, cytokine production, stronger activation of HIF-1a, MAPK, NFκB, and mTOR signaling (with the activation of MAPK being located upstream of mTOR signaling), and increased glucose metabolism. In BMDM and CD4 + T cell co-cultures, rFlaA:Betv1 stimulation significantly suppressed Bet v 1-induced IL-5 and IL-13 secretion while inducing IFN-γ production (summarized in Figure 9). Therefore, macrophages contribute to the strong immune-modulating effects of rFlaA:Betv1 previously observed in vivo [15] and should be considered important target cells contributing to the re-establishment of allergen tolerance in AIT.
Finally, our results suggest that both the fusion of antigens to immune-activating molecules and the resulting structural alterations in the fused antigen can increase its immunogenicity, resulting in the induction of distinct immune responses. Therefore, such adjuvant:antigen fusion proteins likely are valuable tools to induce distinct (in our case balanced Th1) immune responses against antigens with otherwise either low immunogenicity or a high risk of triggering potentially harmful Th2 responses.