Female C57BL/6J mice were obtained from Charles River breeding colony (France); housed in the animal care facilities at IIS-Fundación Jiménez Díaz (Madrid, Spain) under standard laboratory conditions. All animal protocols and procedures were approved by the Institutional Animal Care and Use Committee from Community of Madrid (Ref. PROEX 392/15) and were conducted in compliance with current legislation (European Union Directive 2010/63/EU).

Mice (6–8-week-old) were administered intranasally with recombinant Alt a 1 (12) (50 μg), Complex (50 μg; 1:4 for Alt a 1:quercetin ratio) or saline buffer on days 1, 3, 5, 8 and 11. Control groups of mice that received albumin (BSA or OVA; 50 μg/dose) intranasally were included. Mice were slightly anaesthetized with 3% isoflurane and O₂ at 1L/min (Abbott Laboratories, North Chicago, IL), and once unresponsive but breathing comfortably, 50 μl of sample solution were directly applied on the nostrils. Mice were sacrificed on day 12.

To characterize the Alt a 1 traffic, non-polarized Calu-3 cells (ATCC; HTB-54; USA) were grown on coverslips until reached a 70% of confluency. Then, they were incubated with Alt a 1 at 37°C, 5%CO2 for different times: 5 min for early endosome assay, and 30 min for recycling endosome assays. After incubation, cells were washed, fixed with 4% PFA for 10 min and permeabilized with PBS-Triton X100. Cells were blocked and then they were incubated with anti-Alt a 1 antibody and the proper anti-endosome marker (all antibodies listed in Supplementary Table 1) for 1h at RT. After using Alexa-conjugated secondary antibodies, cells were stained with DAPI and mounted with ProLong Gold over slides. Images were obtained with a Zeiss LSM 880 confocal microscope, using 405, 488 and 633 laser excitations and 63X amplification.

The recycling pathway was inhibited by adding 100 mM endosidin-2 (ES-2; Sigma-Aldrich) 1h prior to adding Alt a 1 to Calu-3 monolayers. Alt a 1 accumulation was evaluated by immunofluorescence following the earlier described protocol.

Silica nanospheres (80 nm carboxyl; nanoComposix, California, USA) were conjugated with TLR 4 (Peprotech, New Jersey, USA) in 50 mM MES buffer pH 4.5 containing 0.01% Tween-20 overnight at RT. The functionalized nanoparticles were blocked with 50 mM Tris-HCl pH 8 and casein blocking buffer for overnight at 4°C. For pull-down assays, nanoparticles were incubated with Alt a 1 (0.1 μg/μL) or Complex (0.1 μg/μl) for 2h at room temperature. After centrifugation, pellets were extensively washed and resuspended in denaturing Laemmli buffer. Samples were separated into a 15% SDS‐PAGE and the biding of Alt a 1 or Complex was confirmed by western blot using a specific anti-Alt a 1 antibody. Negative controls of non-functionalized nanoparticles incubated with Alt a 1 or functionalized nanoparticles without Alt a 1 were included.

Intact lung halves (neither treated nor paraffined) from healthy C57BL6/J mice were incubated with Alt a 1 (25 μg) or saline (PBS) for 10 min at 37°C in RPMI 1640 medium (Invitrogen, USA) containing 2 mM L-glutamine, 25 mM HEPES, 10% heat-inactivated fetal bovine serum, and 1% Pen- Strep. Then, lungs were fixed with 4% PFA for 16 h and cut into thin slices. After blocking, lung slices were stained with specific antibodies (listed in Supplementary Table 1) for 1h at RT, followed by secondary antibody conjugated with Alexa Fluor Dyes. Nuclei were stained with DAPI, and negative controls of anti IgG-Alexa antibodies were included.

The transfected cell line THP1-XBlue-CD14™ (In vivogen, France) derived from the human monocytic THP-1 cell line was used according to the manufacturer’s instructions. Briefly, 1.10 (6) cells/ml were stimulated with Alt a 1 or Complex (5 μg/ml) for 16 h, and, the NF-κB pathway activation was monitored by the detection of the secreted embryonic alkaline phosphatase (SEAP) in the supernatants. For the inhibition tests, cells were incubated with anti-TLR4 antibodies (5 μg/ml; Invitrogen) or an inhibitor of myeloid differentiation protein 2 (10 μM; MedChemExpress, USA) for 15 min prior to the stimulation. In addition, the presence of TNFα and IFNγ was detected in the supernatants from Alt a 1-induced THP1 cells using ELISA kits (ImmunoTools GmbH, Germany), and the results were normalized by the non-stimulated values.

To obtain human macrophages-like cells, THP-1 cells were treated with a final concentration of 100 nM Phorbol 12-Myristate 13-Acetate (PMA; Sigma Aldrich, USA) in RPMI medium for 48h. THP1-derived macrophages were cultured in coverslips and, once they reach a stable morphological state, cells were incubated with Alt a 1 for 2 min at 37°C. Cells were immediately fixed with 4%PFA and TLR4 and Alt a 1-Alexa 647 antibodies were used for an immunofluorescence assay. Nuclei were stained with DAPI and a negative control of secondary antibodies was included. Images were obtained with a Zeiss LSM 880 confocal microscope, using 405, 488 and 633nm laser excitations.

Structural comparison between Alt a 1 and myeloid differentiation factor 2 (MD2) was compute using three structural alignment methods implemented on the protein data bank website: FATCAT (13), CE (14) and TM-Align (15). Moreover, it also was analyzed using MMLigner (16), a statistical inference of protein structural alignments and the presence of concepts in the Alt a 1 structure was identified using Procodic (17). The PDB entries 3V0R (18) and 3FXI (19) were used for Alt a 1 and MD2 structure respectively. In order to analyze the interaction between TLR4 and Alt a 1, the initial geometry was obtained with blind docking calculations using ZDOCK (20), and ColabFold (21). Among the models obtained, the initial geometry was chosen based on the similarity of the Alt a 1 localization with respect to MD2 in the TLR4-MD2 crystal structure (PDB: 3FXI).

Alt a 1-TLR4 molecular system and Alt a 1- TLR4 tetramer system was explored with all-atom 50 and 100 ns Molecular Dynamics (MD) simulations using the CHARMM3.1 force field and the multicore CUDA version of NAMD 2.13 in the Tesla V100 GPU of the high-performance computing CBGP. The system was prepared with CHARMM-GUI (22, 23).Periodic solvation boxes with 14 Å spacing in all dimensions and TIP3P (24) water model were used and Na+ and Cl- ions added to counter total charges and set 0.150 M salt concentration. Initial geometries were minimized at 5000 conjugate- gradient optimization steps and water was then equilibrated at 298 K and 1 atm for 100 ps at 2 fs time steps. Production runs were performed during 100 ns simulation time at 2 fs timesteps in the NPT ensemble at 1 atm and 298 K with Langevin dynamics for T control and Nose-Hoover Langevin piston method for P control. Trajectories were processed and analyzed with VMD 1.9.3 (25).The dissociation constant of the TLR4-Alt a 1 and TLR4-MD2 complexes were calculated using Prodigy web server (26, 27).

Molecular graphics were prepared and rendered with UCSF Chimera v1.16 (28). Protein Data Bank the was also analyzed using MMLigner (16). Structures, Alt a 1-TLR4 and MD2-TLR4 tetramer systems were explored with all-atom 100 ns Molecular Dynamics (MD) simulations using the CHARMM 3.1 force field and the multicore CUDA version of NAMD 2.13 in the Tesla V100 GPU of the high-performance computing CBGP. Periodic solvation boxes and water model were used, and Na+ and Cl- ions were added to counter total charges and set 0.150 M salt concentration. Initial geometries were minimized at 5000 conjugate-gradient optimization steps and water was then equilibrated at 298 K and 1 atm for 100 ps at 2 fs timesteps. Production runs were performed during 100 ns simulation time at 2 fs timesteps (50 million steps per simulation) in the NPT ensemble at 1 atm and 298 K with Langevin dynamics for T control and Nos-Hoover Langevin piston method for P control (28).

Statistically significant differences were analyzed by Prism9 (GraphPad Software Inc., La Jolla, CA, USA) using Kruskal-Wallis test with Dunn’s correction for multiple comparisons and P-values < 0.05 were considered significant.

Intranasal exposure of mice to Alt a 1 (Figure 1A), in the presence (Complex) or absence (Alt a 1) of its ligand (a quercetin-like molecule), elicited allergic airway inflammation. No differences could be observed as a result of the presence of the ligand. Alt a 1 sensitization was confirmed by measuring specific humoral response (Figures 1B–D). This response was only induced by the presence of the allergen and not because of putative environmental pollution, as suggested by the BSA-treated group lacking specific IgE (Figure 1E). Moreover, we observed differences in the respiratory capacity between mice with Alt a 1-induced asthma and controls (Figure 1F), the first ones also showing severe phenotypic changes in the lung tissue such as sub-epithelial fibrosis, cell infiltration and goblet cell hyperplasia (Figure 1G).

The cell infiltration was mostly due to CD45+ cells, especially macrophages and eosinophils (CD45+SiglecF+) and neutrophils (CD45+ F4/80- Ly6G+ SiglecF-) (Figure 2A). Moreover, in the case of the bronchoalveolar lavage fluid (BALF), the high infiltration levels of cells in asthmatic mice was mostly due to eosinophils and AM, (Siglec-F+, F4/80+; Figure 2B). The mediator response was an inflammatory remodeling response. When the cytokines present in BALF were quantified (Figure 2C), they were mostly inflammatory cytokines (such as TIMP-1 or IL1ra) or chemokines inducing monocyte infiltration (such as CCL2, 12 or CXCL12).

The asthmatic phenotype observed can only be due to the presence of Alt a 1. When another allergen was used as OVA following the same pattern as for the Alternaria allergen, no humoral response, no cellular infiltration in the BALF and no alveolar remodeling could be seen (Figures 3A–D).

The asthmatic phenotype of Alt a 1-sensitized mice was characterized by analyzing the expression of asthma markers such as ORMDL-3 (29) and caspase-1 (30). The presence of both markers was increased in the lungs of asthmatic mice after Alt a 1-sensitization as can be seen in Figure 4A. The characterization of the asthmatic profile was completed by transcriptomic analysis study comparing asthmatic versus healthy mice. In this way, 42 genes were found up-regulated (> 4-fold), most of which are related to eosinophilic asthma. The majority of those genes (n=27) corresponds to T2 cytokines and alarmins (i.e., IL25, IL33, IL13); chemokines (i.e., IL5); their receptors (i.e. IL4ra, IL5ra) (Figure 4B); or genes related to lung impairment (i.e., Muc5a, Ccll2 or Ccll3) and mast cell functions (i.e., Fcer1a, Tpsb2 or Kitl) (Figure 4C).

To study how Alt a 1 can pass through the epithelial barrier, an in vitro bronchial epithelium model was developed using the human Calu-3 cell line. According to previous data, Alt a 1 can interact with epithelial cells through the SLC22A17 receptor, however the percentage of transported protein is too low to explain the recognition by APCs11. Studying the Alt a 1 internalization into epithelial cells, no co-localization was observed between Alt a 1 and EEA1 (early endosome antigen-1), an early endosome marker, suggesting that protein follows an EEA1-independent pathway. In contrast, Alt a 1 co-localized with Rab 11, a marker of the endosomal-recycling pathway (Figure 5A). This result was confirmed through the inhibition with endosidin-2 (ES-2), an exocytosis and endosomal recycling inhibitor (Figure 5B). Thus, Alt a 1 could be recycled following the same route as in the case of lipocalin-2, another ligand of SLC22A17 (Figure 5C) (31).

Since the recycling of Alt a 1 might be partially preventing the apical-to-basal transport of the protein through the epithelial barrier, and considering that the protein does not alter the epithelial permeability in vitro, the mechanism through which Alt a 1 reaches APCs on the basolateral side remains unclear. One possibility is the recognition of Alt a 1 by AM, which could directly introduce the allergen in the alveoli triggering a pro-inflammatory response, as previously described6. Supporting this hypothesis, co-localization of Alt a 1 with TLR4 in mice lung F4/80+ cells (Figure 6A) as well as in THP1-derived macrophages (Figure 6B) was observed in immunofluorescence assays. This recognition was also confirmed by pull down-binding assay, using TLR-nanoparticles incubated with Alt a 1 (Figure 6C). When Alt a 1 was cultivated with a macrophage cell line, THP1-XBlue-CD14+, the interaction between TLR4 and Alt a 1 was inhibited by adding specific antibodies or specific MD2 inhibitor (32) (Figure 6D).

Based on these results, the structural similarity between Alt a 1 and MD2 was analyzed. Both are small proteins of 15 and 18.5 KDa respectively with a structure characterized by the presence of beta sheets. Although the scores obtained in the similarity analysis do not reveal that they belong to the same protein family, a structural relationship between both can be observed (Figure 7A).

To explore whether such structural similarity might be sufficient to complex with TLR4, the interaction between Alt a 1 and TLR4 was explored by means of molecular dynamics simulations using as a reference the interaction between TLR4 and MD2. The results obtained, reveal that the Alt a 1-TLR4 complex remains stable along the simulation (100 ns), as shown by the stabilization of the RMSD values (Figure 7B). This complex is maintained by the presence of more than 18 hydrogen bonds (Figure 7C). In addition, the distances between the Glu 586 and the Asp 325, in the C terminal region of the TLR4 ectodomain, between each protein of the homodimer were analyzed to know the behavior along the dynamics simulations. It can be seen how in the case of TLR4 with Alt a 1, the distance between the residues is more similar to the crystallographic structure of TLR-MD2-LPS than in the case of TLR4-MD2, in which a high increase could be observed (Figure 7D). All these data suggest that Alt a 1 could be bound with high affinity to TLR4, in absence of MD2, promoting the dimerization of the receptor.