Chimeric allergen receptor regulatory T cells suppress birch pollen allergic airway inflammation

Asthma affects over 300 million people worldwide and remains a deadly disease if insufficiently treated, representing a high burden on individuals, caregivers, and healthcare systems (Papi et al., 2018, Global Initiative for Asthma, 2026). Allergic asthma is driven by an exacerbated type 2 immune response, characterized by the overproduction of IL-4, IL-5, and IL-13 by Th2 cells. These cytokines promote IgE class switching, eosinophil recruitment, and mast cell activation resulting in airway hyperresponsiveness and airway mucus plugging, the principal cause of death in asthma (Maddox and Schwartz, 2002). Its global economic burden is estimated at $80–100 billion per year, with severe cases, which represent 5–10% of asthma patients (Wang et al., 2020) requiring biologic therapies reaching $20,000–40,000 per patient annually (Grønseth and Jansson, 2014; Wang et al., 2023). Despite the approval of several anti-cytokine biologics, allergen immunotherapy (AIT) remains the only disease-modifying therapy (Brusselle and Koppelman, 2022; Cox et al., 2011; Israel and Reddel, 2017).

A major cause of allergic rhinitis and asthma is related to birch sensitization affecting 8–16% of the European population (Biedermann et al., 2019). Birch pollen cross-reacts with a large family of trees including alder, hazel, oak, hornbeam, chestnut, and beech. This birch homologous group shares highly cross-reactive allergens derived from the pathogenesis-related 10 protein (PR-10) family, among which the birch allergen Bet v1 is the immunodominant and most abundant allergenic protein (Schenk et al., 2011). While AIT for birch pollen–associated rhinitis and asthma has been shown to be effective (Wahn et al., 2019), it remains contraindicated in patients with severe and uncontrolled asthma and is poorly effective in polysensitized patients (Cox et al., 2011). This highlights the unmet need for new, safe, and durable treatments for restoring allergen tolerance in severe allergic asthma.

Regulatory T cells (Tregs) are crucial for suppressing type 2 inflammation through multiple modalities including IL-2 consumption, production of IL-10 and TGF-β, and downregulation of dendritic cell immunogenic activities (Akbari et al., 2003). Importantly, Tregs can be expanded ex vivo and reinfused with multiple clinical trials evaluating their potential in autoimmune and inflammatory disorders (Raffin et al., 2020). However, Treg therapy has shown only limited efficacy, which has been mostly attributed to the lack of antigen specificity (Baeten et al., 2022). To overcome this limitation, efforts have focused on redirecting Treg specificity by engineering synthetic receptors such as chimeric antigen receptors (CARs) (Baeten et al., 2022).

Herein, we hypothesized that Tregs can be armed with a chimeric allergen receptor (CAlleR) to target allergens such as Bet v1 to reset tolerance against birch pollen–associated allergic diseases. We identified and characterized four novel anti–birch-specific antibodies and generated single-chain variable fragments (scFvs) fused to a CD28-ζ signaling domain. We demonstrated allergen-specific response of CAlleRs. To decipher the physiological mechanisms by which CAlleR activation occurs, we showed that soluble allergens can induce maximal activity of CAlleRs when stabilized by a noncompetitive allergen-specific antibody in a Fcγ receptor (FcγR)–dependent manner. Finally, CAlleR Tregs preserved lung function in birch pollen–sensitized mice.

CD19⁺IgM⁻IgG⁺ B cells from a birch-allergic donor were sorted and immortalized for single-cell plating (Fig. S1 A). Bet v1 specificity for 5,760 IgG and 20 IgE clones was assessed by ELISA. We could identify four IgG and three IgE Bet v1–specific clones after Sanger sequencing of the heavy variable and light variable gene pairs (Fig. 1, A and B; and Fig. S1 B). Two of the four IgG clones (monoclonal antibody [mAb] 5 and mAb 8) shared the same V gene, IGHV5-51 for the heavy and IGKV1-39 for the light but had different complementarity-determining region 3 (CDRH3) (Fig. 1 B). Similarly, two of the three IgE clones (eB13 and eC5) shared the same V genes for the heavy (IGHV3-7) and the light (IGKV2-28) chains with different CDRH3 (Fig. S1 B). To validate the antibody specificity, we produced all selected mAbs, as well as three previously reported control anti-Bet v1 mAbs (REGN5713, REGN5714, and REGN5715) (Atanasio et al., 2022). Bet v1–specific IgG (5, 8, 10, 11) and IgE (eB13, eB20, and eC5) clones showed similar half maximal effective concentration (Fig. 1, B and C; and Fig. S1, B and C).

To further evaluate the cross-reactivity of our antibodies against the PR-10 family, we tested the binding specificity to Cor a1 (homology to Bet v1 72%), Mal d1 (homology to Bet v1 58%), and Gly m4 (homology to Bet v1 47%). Interestingly, clones 5 and 8 cross-reacted with Cor a1 and clones 10 and 11 cross-reacted with Mal d1, while the remaining only reacted against Bet v1. Importantly, none of the antibodies bound unrelated allergens (Fig. 1 D and Fig. S1 D). We next assessed their binding affinity to Bet v1 by the biolayer interferometry (BLI) assay. Clones 5, 8, eB13, and eB20 showed higher association and almost no dissociation compared with mAb 10, 11, and eC5 (Fig. 1 E and Fig. S1 E). Similarities between mAbs 5 and 8 and between eC5 and eB13 were further supported by hierarchical clustering of antibody sequences (Fig. S1 G). Considering that the cross-recognition of homologous PR-10 allergens represents a potential advantage, we further selected clones 5, 8, 10, and 11 for the downstream analysis.

We next tested the neutralization capacity of the antibodies to block Bet v1–specific IgE binding from three birch-allergic donors. Clones 5 and 8 showed the strongest blocking capacity individually. Despite the high affinity of mAbs 10 and 11, they showed lower blocking effect (Fig. 1 F). Epitope binning by cross-competition binding to Bet v1 revealed that mAbs 5 and 8 recognized the same epitope as REGN5713, which partially overlapped with that bound by mAb 11. In contrast, mAb 10 competed for a partially overlapping epitope also shared with mAb 11 (Fig. 1 G). Among the IgE clones, eC5 inhibited binding of mAbs 5, 8, and 10, and REGN5713. Similarly, eB13 competed with these antibodies and partially with mAb 11. In contrast, eB20 did not compete with the other clones for Bet v1 binding, except with REGN5714 (Fig. S1 H). To confirm that mAbs 8 and 10 recognized distinct regions of Bet v1 and to gain further insight into their epitopes, Bet v1 was complexed with threefold molar excess of Fab08 and Fab10 and analyzed by cryo-electron microscopy (cryo-EM). Reconstruction of the Bet v1-Fab08-Fab10 complex was obtained at a resolution of 4.8 Å, allowing Cα positioning (Fig. 1 H). Bet v1 adopted the canonical fold consisting of a seven-stranded antiparallel β-sheet wrapped around a 25-residue-long C-terminal amphipathic α-helix, as previously described (Gajhede et al., 1996). Fab08 sandwiched the C-terminal α-helix, similarly as REGN5713, with interactions mediated almost exclusively by its heavy chain. In contrast, Fab10 bound a distinct epitope consisting of a loop formed by residues 60–65, which lies within the epitope region recognized by REGN5715 (Fig. 1 I) (Atanasio et al., 2022). Overall, the binding, cross-reactivity, and functional properties of these novel anti-Bet v1 IgG mAbs support their further evaluation as scFvs for the design of CAlleRs.

We next generated scFvs derived from our four anti-Bet v1 IgG mAbs and fused them to a CD28 hinge, transmembrane, and costimulatory domains followed by a CD3ζ intracellular signaling domain. Anti-CD19 CAR was used as a control and mCherry as a reporter (Fig. 2 A). To evaluate the functional activity of the CAlleRs, we first engineered artificial K562 cells with a platelet-derived growth factor (PDGF)–truncated receptor covalently linked to the Bet v1 recombinant protein and a GFP molecule on the N and C terminus, respectively (Fig. 2 B). To validate the specificity of the selected CAlleRs, Jurkat NFAT reporter cell lines were transduced with lentiviruses encoding the CAlleRs and cocultured with K562 Bet v1 for 24 h (Fig. 2C). CAlleR 5 and 8 displayed the highest NFAT activity, followed by CAlleR 11 and CAlleR 10 (Fig. 2 D).

Primary human CAlleR T cells were then generated to validate their functionality in vitro (Fig. 2 E). Similar transduction efficiencies were obtained (prior to calculation of multiplicity of infection [MOI] = 1) (Fig. 2, F and G). Again, CAlleR 5 and 8 showed higher binding to biotinylated Bet v1 compared with CAlleR 11 and 10 (Fig. 2 G), correlating with CD25 and CD71 upregulation (Fig. 2, H–J).

The suppressive capacity of CAlleR Tregs was evaluated by applying a similar editing strategy to freshly isolated CD4⁺CD25⁺CD127^low Tregs (Fig. 3, A and B). After 7 days of culture, the cells expanded 20–40-fold (Fig. 3 C) and maintained a stable Treg phenotype independently of CAlleR expression (Fig. 3 D). Similar transduction was observed among all CAlleRs (Fig. 3 E). We observed a Bet v1–induced upregulation of early activation markers (OX40 and 41BB) in CAlleR Tregs (Fig. 3 F). Interestingly, the higher activation rates observed with CAlleR 5 and 8 Tregs were consistent with those obtained with T cells (Fig. 2, H and I; and Fig. 3 F). To evaluate their suppressive function, CFSE-labeled CAlleR 8 or CAlleR 11 T cells were cocultured with K562 Bet v1 and different ratios of CAlleR or polyclonally expanded Tregs (Fig. 3, G and H). Overall, CAlleR Tregs exhibited greater suppressive capacity than polyclonal Tregs correlating with the in vitro activation levels. CAlleR 8 Tregs exhibited the highest suppressive capacity for both CAlleR 8 and 11 T cells. We therefore selected this candidate for the in vivo studies.

Since CAR signaling for soluble ligands relies on ligand-mediated dimerization and eukaryotic cells do not constitutively express allergens (Chang et al., 2018), we hypothesized that anti-Bet v1 antibodies can stabilize soluble antigens and contribute to CAlleR activation. Indeed, soluble Bet v1 alone was insufficient to activate CAlleR T cells (Fig. 4 A). Yet, in the presence of noncompetitive mAbs we observed a modest upregulation of activation markers (Fig. 4 A) in line with the competition binding assays (Fig. 1, G–I). To further assess the antibody-dependent activation, we examined the effect of low-, medium-, and high-affinity FcγR (CD16, CD32, and CD64, respectively) by generating specific K562 cell lines (Fig. S2, A and B). The activation of CAlleR T cells was FcγR-dependent and required the stabilization of the Bet v1 by noncompetitive birch-specific antibodies (Fig. 4 A). Importantly, we observed an antibody dose-dependent activation of the CAlleR 8 in the presence of mAb 10 (Fig. 4, B and C; and Fig. S2 C). Furthermore, two noncompetitive scFvs are sufficient to stabilize and crosslink the allergen to induce CAlleR T cell activation in the presence of soluble Bet v1 (Fig. S3, A–C). The capacity of noncompetitive Bet v1–specific antibodies to induce OX40 and 4-1BB upregulation in CAlleR Tregs was further confirmed (Fig. 4 D) and correlated with their trogocytosis capacities (Fig. 4 E). The suppressive capacity of CAlleR 8 Tregs was FcγR and antibody concentration–dependent and better than polyclonal Tregs in all conditions (Fig. 4, F and G). Interestingly, the suppression was less efficient against CAlleR 8 T cells stimulated with mAb 10 bound to CD64 compared with CD32, suggesting more resistance to Treg when effector T cells (Teff) are strongly activated (Fig. S2 C).

The in vivo functionality of the CAlleR was first validated by engineering murine Teff and monitoring CFSE dilution. CAlleR 8 Teff were infused in birch pollen extract (BPE)–sensitized mice (Fig. 5, A–D). Their presence and proliferation were compared in the lung, mediastinal lymph node (mLN), cervical lymph node, and the spleen of PBS and BPE-sensitized mice. CAlleR 8 and polyclonal Teff were detected in all organs (Fig. 5 E). Yet, CAlleR 8 Teff proliferated significantly more than their polyclonal counterpart in a birch-dependent manner, predominantly in the lungs and draining mLNs (Fig. 5 F). To confirm CAlleR 8 Teff proliferation in response to birch-specific immunoglobulins in a FcγR-dependent manner, we purified bone marrow–derived dendritic cells (BMDCs) expressing high levels of FcγR (Guilliams et al., 2014). We then investigated whether BPE-exposed mice produced specific immunoglobulins capable of mediating CAlleR 8 T cell activation via the FcγR on BMDCs (Fig. 5 G). After the sensitization protocol, high levels of anti-Bet v1 IgG1 were detected in allergic mouse serum (Fig. 5 H), which induced CAlleR 8 T cell activation when BMDCs and the allergen were present (Fig. 5, I and J). These findings demonstrate that CAlleR 8 Teff proliferate in allergen-exposed organs in a birch-specific antibody–dependent manner. A similar experimental design was applied using peripheral blood mononuclear cells (PBMCs) and serum samples from birch-allergic donors (Fig. 5 K). However, the significantly lower levels of anti-Bet v1 IgG in donor serum compared with sensitized mice (Fig. 5 L) were likely insufficient to trigger CAlleR 8 T cell activation in autologous coculture with decomplemented serum and CD64-expressing K562 cells (Fig. 5 M).

Considering the risks associated with CAR T cell therapies, we generated CAlleR murine Tregs expressing either the CAlleR 8 or a control anti-HLA-A2 CAR (Wagner et al., 2022) with Thy1.1 as a reporter (Fig. 6 A). After 7 days of culture, phenotype and transduction efficiency were evaluated (Fig. 6 B) prior to injection to BPE-exposed mice (Fig. 6 C). Treg-treated mice had lower cell influx in the bronchoalveolar lavage fluid (BALF) than allergic mice indistinctively of their specificity (Fig. 6 D). Importantly, CAlleR 8 Tregs significantly reduced the eosinophil percentage in the BALF (Fig. 6 E). Additionally, Treg infusion correlated with increased IL-10 levels in the BALF (Fig. 6 F). In the lungs of Treg-treated mice, lower levels of eosinophils (CD45⁺CD11c⁻CD11b⁺Ly6G⁻Ly6C⁻SiglecF⁺) were detected (Fig. 6 G and Fig. S4 A). Notably, the expression of SiglecF, a marker for eosinophilic activation, was also decreased on these cells (Fig. 6 H and Fig. S4 A). These findings coincided with reduced birch-specific IgG and total IgE levels in the serum of these mice (Fig. S4, B and C). However, dendritic cell (CD11c⁺CD11b⁺F4/80⁻MHC-II⁺) infiltration in the lungs was not reduced (Fig. S4 D). Although no overall differences were observed in the proportions of T cell populations in the lungs and mLNs (Fig. S4, E and F), Treg-treated mice exhibited reduced CD4⁺ T cell activation, as indicated by reduced CD44 expression (Fig. S4 E), along with diminished production of IL-4, IL-13, and IL-17 (Fig. 6 I and Fig. S4 G). Lung mucus production was quantified using periodic acid–Schiff (PAS) staining. Mice treated with CAlleR 8 Tregs showed fewer PAS⁺ airways than the other groups (Fig. 6, J and K). This finding was supported by improved lung function with CAlleR 8 Treg–treated mice showing reduced airway hyperreactivity, as measured by lung stiffness (H) and tissue resistance (G) after methacholine challenge (Fig. 6 L). Overall, these data demonstrate the protective nature of Treg in a preclinical mouse model of allergic airway inflammation that can be further enhanced with CAlleRs.

We next evaluated whether Tregs can prevent allergic disease development. To this end, engineered Tregs were adoptively transferred to mice prior to their first exposure to BPE (Fig. 7 A). In this context, only the CAlleR 8 Tregs reduced the cellular influx in the BALF, with a lower proportion of eosinophils and lymphocytes (Fig. 7, B and C). Increased IL-10 was also detected in the BALF of CAlleR 8 Treg–treated mice (Fig. 7 D). In the lung, no significant differences were observed in the proportion of neutrophils (CD45⁺CD11c⁻CD11b⁺Ly6G⁺), monocytes (CD45⁺CD11c⁻CD11b⁺Ly6G⁻Ly6C⁺), and eosinophils (CD45⁺CD11c⁻CD11b⁺Ly6G⁻Ly6C⁻SiglecF⁺) (Fig. 7 E). Yet, SiglecF high eosinophils were significantly less prevalent in mice treated with CAlleR 8 (Fig. 7 F). Although in this model the anti-Bet v1 IgG1 production remained unaffected by Treg transfer (Fig. S5 A), CAlleR 8 Tregs reduced total IgE levels (Fig. S5 B). CAlleR 8 Tregs did not significantly affect the dendritic cell (CD11c⁺CD11b⁺F4/80⁻MHC-II⁺) infiltration in the lungs (Fig. S5 C). Similarly, Treg prophylactic treatment had no effect in the T cell compartment, neither on the proportions of the different populations nor on their activation status in the lungs and the mLNs (Fig. S5, D and E). However, significantly less IL-4 production by CD4⁺ T cells was observed (Fig. 7 G). Importantly, mice treated with CAlleR 8 Tregs exhibited reduced mucus production (Fig. 7, H and I), a finding associated with significantly improved lung function with reduced airway hyperreactivity, as measured by reduced airway (Rn) and tissue (G) resistance, as well as lung stiffness (H) after methacholine challenge (Fig. 7 J). These data demonstrate that CAlleR 8 Tregs can prevent the inflammatory response of birch pollen–induced allergic asthma.

To further investigate the mechanism of CAlleR Treg in vivo, we engineered the cells with a GFP reporter (Fig. 8 A). After in vitro expansion, transduction and phenotype of the engineered Tregs were confirmed (Fig. 8 B) and cells were adoptively transferred to mice, which were then exposed to BPE for four consecutive days (Fig. 8 C). In mice exposed to BPE, the number of CAlleR 8 Tregs was significantly higher in the lungs and draining lymph nodes (mLN) compared with PBS control mice (Fig. 8, D and E). Importantly, CAlleR 8 Tregs made significantly more contacts with CD11c⁺ APCs present in the mLN than unspecific A2-CAR Tregs (Fig. 8 F). This difference was not observed in the lung and spleen of the mice (Fig. 8 F). Thus, the CAlleR enabled specific migration and interactions of the Tregs with CD11c⁺ APCs present in the mLN.

Herein, we provide proof-of-concept and preclinical evidence that CAlleR Treg redirected against Bet v1 can downmodulate birch pollen–induced allergic airway inflammation. While polyclonal, allergen-unspecific Tregs demonstrated partial protective effects under inflammatory conditions, they were insufficient to prevent disease onset in the absence of inflammation. Importantly, we identified a new mechanism underlying the activation of CAR T cells targeting soluble antigens: engagement through high-affinity FcγRs recruiting antibodies, which are specific for a noncompetitive epitope that preserves the binding site of the CAR. This mechanism opens new avenues not only for rewiring synthetic receptors against any soluble antigens including autoantigens for therapeutic intervention but also for delineating a more global pathway for antigen cross-presentation in allergies. Several attempts have been made to develop CAR T therapies targeting soluble factors, such as mesothelin ligand (Lanitis et al., 2012), carcinoembryonic antigen (Nolan et al., 1999), factor VIII (Fu et al., 2020), Lewis-Y carbohydrate antigen (Westwood et al., 2009), BCMA (Carpenter et al., 2013), CD30 (Hombach et al., 1998), or insulin (Tenspolde et al., 2019) with limited success. In fact, soluble antigens in their monomeric form, as for nominal Bet v1 antigen, consistently failed to induce CAR T cell activation (Chang et al., 2018). In contrast, dimeric soluble ligands, such as TGF-β, can cluster CARs together and mediate T cell responsiveness, an actin-dependent dynamic process relying on mechanical forces between the targeted antigen and the CAR (Chang et al., 2018; Qiu et al., 2024). In this study, we discovered that CAlleR T cells can respond to soluble Bet v1 in a FcγR-dependent manner when specific antibodies not interfering with the CAR (noncompetitive) are present. As high titer of allergen-specific IgGs was detectable in birch-sensitized mice, this mechanism of allergen presentation could indeed contribute in the recruitment and activation of CAlleR Tregs. One could also speculate on the existence of a similar cross-presentation pathway with allergen-specific B cell receptor. Thus, the relevance of this mechanism warrants further investigation, which should also include a prospective and longitudinal monitoring of sensitized patients during on- and off-pollen period season.

Our data indicated a preferential migration of CAlleR Tregs to the mLNs and the lungs, reinforcing initial observations that redirecting the specificity of Tregs is sufficient to retain them in various tissues, such as the gut (Porret et al., 2025) or an HLA-A2 islet graft (Muller et al., 2021a). Yet, in inflammatory conditions, i.e., during the second round of allergen sensitization, we also observed a functional activity of the nonspecific CAR-edited Tregs. While the HLA-A2 CAR remains inactive in HLA-A2–negative control mice, thus effectively ruling out off-target activity (Wagner et al., 2022), its enhanced recruitment in the lung may be associated with the pulmonary passage following cell infusion (Fischer et al., 2009), which, in combination with pro-inflammatory chemokines, promotes homing and persistence of the cells in the lungs. Imprinting of Tregs by lung dendritic cells could favor CCR4 upregulation and thereby a more efficient trafficking to the lungs (Mikhak et al., 2013). Afshar et al. demonstrated that CCR7 on Treg is required to suppress allergic airway inflammation during the sensitization phase, whereas CCR4 is essential for suppressing inflammation during the effector phase (Afshar et al., 2013). Future studies should better characterize the interplay between the CAlleR and the chemokine receptors in the trafficking dynamics between the mLN and the lung.

Our data build on previous evidence suggesting that the transfer of Tregs can modulate allergic airway inflammation in vivo (Kearley et al., 2005; Xu et al., 2012). The main advantage of Tregs over other cell therapy modalities, such as CAR T cells redirected against IL-5 and secreting IL-4/13 muteins (Jin et al., 2024), is their safety profile, as they lack proliferative capacities and produce considerably less pro-inflammatory cytokines (Esensten et al., 2018). Our findings show that CAlleR Tregs reduced levels of IL-4 in CD4⁺ T cells, which could explain the decrease of eosinophils in the BALF as IL-4 mediates eosinophils’ transendothelial migration (Moser et al., 1992). Beyond their immunoregulatory functions, Tregs also contribute directly to tissue repair and homeostasis through amphiregulin-dependent mechanisms, an additional remarkable advantage over more conventional T cell therapies (Arpaia et al., 2015). Yet, bona fide self-antigen–specific Tregs can lose Foxp3 expression during an inflammatory autoimmune response (Bailey-Bucktrout et al., 2013) or upon repetitive stimulation (Hoffmann et al., 2009). Thus, future studies should evaluate the persistence and stability of the CAlleR Tregs over time and define the optimal modalities for implementing such approach, e.g., outside the pollen season or as an adjunct to a desensitization protocol to reduce local inflammation.

The present study has several limitations. First, we focused on birch pollen–related allergic airway inflammation. However, patients with severe allergic asthma often suffer from multiple sensitizations against PR-10–unrelated proteins, such as profilins or polcalcins (Cacheiro-Llaguno et al., 2024; Ruiz-Hornillos et al., 2020). Therefore, future studies should evaluate the potential cross-reactive and bystander suppression of birch pollen–specific CAlleR Tregs and their capacity to suppress PR-10–related and unrelated memory T cells, respectively. Second, the in vivo effects of the CAlleR Tregs remain modest compared with other cell therapy products, such as IL-5–targeting CAR T cells (Jin et al., 2024). While the security profile of Treg-based therapy is certainly higher, optimized CAlleR designs combining several complementary allergen–specific scFvs may be necessary to improve their responsiveness to soluble allergens. Importantly, future studies should evaluate the response of CAlleR Tregs from allergic patients during and outside the pollen seasons. Third, redirecting Treg specificity toward perennial allergens such as house dust mite may be also relevant considering the challenges associated with allergen avoidance. Accordingly, future studies should investigate the suppressive capacity of CAlleR Tregs in polysensitized mouse models. In parallel, it will be critical to assess the phenotypic stability and functional activities of CAlleR Tregs under repeated allergen stimulation as we have not investigated their persistence beyond 5 days after injection. Also, we did not assess the suppressive capacity of Tregs on allergen-specific B cells, which may also contribute to allergen mobilization through their BCRs. Finally, future studies should evaluate whether FcγR-dependent allergen cross-presentation is also relevant in other cell types, including B cells, as well as basophils and mast cells.

In conclusion, we demonstrate that CAlleR Tregs can effectively prevent and downmodulate birch pollen–induced allergic airway inflammation. Our data provide compelling evidence for the role of noncompetitive “allosteric” antibodies in enhancing the action and homing of Tregs, contributing to the restoration of immune tolerance to soluble allergens. Future work should evaluate whether such approach could also be suitable to restore tolerance against food allergies.

Deidentified human peripheral blood from healthy donors was ordered at the Swiss Transfusion Center as buffy coats. For B cell sorting, PBMCs were isolated from a birch-allergic female donor enrolled in the Immuno-IgE study. The study was approved by the Institutional Review Board of the Lausanne University Hospital (2020-02798; CER-VD, Switzerland). Written and informed consent was obtained prior to sample collection. Ficoll-Paque (Cat. #17144003; Cytiva) density gradient centrifugation was used to isolate PBMCs.

B cells were isolated from PBMCs with EasySep Human B Cell Isolation Kit according to the manufacturer’s instructions (Cat. #17954; StemCell Technologies) and labeled with LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Cat. #L34957; Invitrogen), mouse anti-human CD19 APC (Cat. #555415; BD Biosciences; clone SJ25C1), mouse anti-human IgM-PECy7 (Cat. #314532; BioLegend; clone MHM-88), mouse anti-human IgG AF488 (Cat. #410706; BioLegend; clone M1310G05), mouse anti-human IgE PE (Cat. #325506; BioLegend; clone MHE-18), mouse anti-human CD27 (Cat. #646851; BD Biosciences; clone HB7). Viable CD19⁺CD27⁺IgM⁻IgG⁺IgE⁻ B cells were sorted with BD FACSAria II cell sorter (BD Biosciences). B cell immortalization was performed as previously described (Traggiai et al., 2004). Briefly, B cells were incubated for 3 h at 37°C in complete DMEM (Cat. #A4192102; Thermo Fisher Scientific) supplemented with 10% FCS, 1% penicillin–streptomycin, 1% MEM nonessential amino acids (Cat. #11140050; Thermo Fisher Scientific), 1% L-glutamine, 1% Na-pyruvate, 30 μg/ml transferrin (HOLO) (Cat. #7542-100; Biovision), 1% kanamycin, 55 μM 2-mercaptoethanol (Cat. #21985-023; Gibco), 2 μg/ml CL264 TLR-7 agonist (Cat. #tlrl-c264e-5; InvivoGen), and 40% Epstein-Barr virus obtained from human gammaherpesvirus 4 cell’s supernatant (Cat. #VR-1492; ATCC). B cells were plated at three cells per well with 26,000 autologous irradiated PBMCs in a 384-well plate. Screening for each well was performed with an indirect Bet v1 ELISA detecting anti-Bet v1 antibodies in the supernatant. Among 5780 clones, we could identify 21 binders, which were further selected for RT-PCR as previously described (Fenwick et al., 2021). Among those 21 clones, 4 IgG and 3 IgE paired heavy and light variable chain productive sequences were obtained with Sanger sequencing (Fasteris). PCR amplification with custom primers (Microsynth AG) was performed to extract the variable chain sequences of the antibodies and for subcloning into AbVec2.0-IGHG1, AbVec1.1-IGKC, or AbVec1.1-IGLC2-XhoI vectors (#80795, #80796, and #99575; Addgene plasmids, respectively). Resulting plasmids were used for transient transfection of ExpiCHO cells with 1.5 μg/ml of each plasmid. After 7 days of cell culture, supernatants were collected and antibodies were purified using Sartobind Lab protein A columns according to the manufacturer’s instructions (Cat. #93PRAP06HB-12--A; Sartorius AG).

The Bet v1 DNA sequence was codon-optimized for Escherichia coli, synthesized by GenScript, and cloned into a bacterial expression vector, pET29b (#110168; Addgene, RRID:Addgene_110168), with a 10xHis-tag at the C terminus and an AviTag. Recombinant Bet v1 (rBet v1) was expressed in E. coli BL21(DE3) cells by growing cells at 37°C until an OD of 0.6, followed by induction with 0.5 mM isopropyl β-D-1-thiogalactopyranoside overnight at 18°C. The bacterial pellet was harvested for 15 min at 5,000g, and was lysed using sonication in buffer A (20 mM HEPES 7.5, 700 mM NaCl, 10% glycerol). Clarification of the cell’s lysate was performed by centrifugation for 30 min at 30,000g. The protein was purified by nickel affinity, with a HisTrap HP 5 ml column, washed with buffer A, and then eluted on a linear gradient of buffer B (buffer A + 500 mM imidazole, pH 7.5). Eluted fractions of rBet v1 were concentrated and injected onto a Superdex 75 16/600 gel filtration column (GE Healthcare/Cytiva) in 20 mM HEPES 7.5, 250 mM NaCl. After size exclusion, PBS buffer exchange was performed by dialysis and rBet v1 was diluted at 1 mg/ml and stored at −20°C.

Antibody binding toward Bet v1 was assessed by indirect ELISA. NuncSorp plates were coated with 2 μg/ml of rBet v1 diluted in coating buffer (15 mM Na₂CO₃, 34.87 mM NaHCO₃) overnight at 4°C, washed with PBS/Tween 0.05%, and blocked with PBS-BSA 1% for 2 h at room temperature (RT). Serial dilutions of the seven novel mAbs were performed starting at 1 μg/ml and incubated for 2 h. Serum samples from mice were diluted at 1:100 prior to incubation. Plates were then washed, and 1 μg/ml of biotin mouse anti-human IgG antibody (Cat. #555785; BD Pharmingen; clone G18-14, RRID:AB_396120), biotin mouse anti-human IgE antibody (Cat. #325504; BioLegend; clone MHE18, RRID:AB_830849), or 2 μg/ml goat anti-mouse IgG1 cross-adsorbed secondary antibody, Biotin-XX (Cat. #A10519; Life Technologies, RRID:AB_10374477) was used, respectively, for detection and incubated for 1 h at RT. Streptavidin–horseradish peroxidase conjugate (Cat. #554066; BD Pharmingen; 1:1,000 dilution) was added and incubated for 1 h at RT. Tetramethylbenzidine substrate (Cat. #555214; BD Biosciences) was added for 20 min, and the reaction was stopped with 2N sulfuric acid. Absorbance was measured on a spectrophotometer at 450 nm (630 nm reference). Binding of mAbs to PR-10 allergens (rMal d1 [Bühlmann Laboratories AG], rCor a1 [Cat. #RE-CA101-1; Inbio], and rGly m4 [Cat. #RP-GM4-1; Inbio] and to other non-PR-10 allergens [kindly provided by Dr. Régine Audran [Lausanne University Hospital, Lausanne, Switzerland] and Dr. Craig Fenwick [Lausanne University Hospital, Lausanne, Switzerland] [rag weed, grass, Derp2, bee venom]) was assessed using the same protocol, coating PR-10 allergens at 2 μg/ml and non-PR-10 allergens at 10 μg/ml. For total quantification of murine IgE, purified goat anti-mouse IgE antibody (Cat. #1110-1; SouthernBiotech, RRID: AB_2794601) was coated at 2 μg/ml in PBS and incubated overnight at 4°C. After washing and blocking, 1:50 dilution of serum samples and twofold serial dilutions of purified mouse IgE for standard curve starting at 2 μg/ml (Cat. #553481; BD Pharmingen, RRID: AB_10050441) were added to the wells. After 2 h of incubation, wells were washed and detection goat anti-mouse IgE-AP antibody was added at 2 μg/ml (Cat. #1110-04; SouthernBiotech, RRID: AB_2794603) and incubated for 2 h. Finally, wells were washed and 4-nitrophenyl phosphate disodium salt hexahydrate (Cat. #N2765-100TAB; Sigma-Aldrich) was diluted in diethanolamine buffer (1M, pH 9.8, Cat. #31590-250G; Sigma-Aldrich), added to the wells, and incubated for at least 5 min prior to absorbance measurement at 450 nm. IL-10 ELISA was performed according to the manufacturer’s instructions (ELISA MAX Standard Set Mouse IL-10, Cat. #431411; BioLegend).

Heavy and light chain amino acid sequences were aligned using ANARCI (Dunbar and Deane, 2016) with the IMGT numbering scheme, and concatenated per antibody. Pairwise sequence similarity was calculated as 1 minus the normalized Hamming distance between concatenated sequences. The resulting distance matrix was hierarchically clustered and visualized as a dendrogram using the ComplexHeatmap package (Gu et al., 2016).

Affinity toward Bet v1 for each antibody was assessed with Gator Prime Core BLI System by using a protein A–coated probe (Cat. #160001; Gator). After binding of the antibodies to the biosensors, tips were dipped into 500 or 300 nM of the Bet v1 protein. The strength of association and dissociation of the Bet v1 to the antibodies was measured.

Covalent coupling of rBet v1 to Luminex beads was performed according to the manufacturer’s instructions with BioPlex Amine Coupling Kit (Cat. #171406001; Bio-Rad). 30-fold molar excess of capture mAbs was combined with rBet v1–coupled beads for 30 min. Biotinylated competitor mAbs were then added to each well and incubated for further 20 min. Biotinylated mAbs bound to rBet v1 were stained with streptavidin-PE (Cat. #554061; BD Pharmingen; 1:1,000 dilution) and analyzed on 200 BioPlex instruments.

Grids were screened for particle presence and ice quality on a TFS Glacios microscope (200 kV), and the best grids were transferred to a TFS Titan Krios G4. Cryo-EM data were collected using a TFS Titan Krios G4 transmission electron microscope, equipped with a Cold-FEG on a Falcon IV detector in electron counting mode. Falcon IV gain references were collected just before data collection. Data were collected using TFS EPU version 2.12.1 utilizing the aberration-free image shift protocol, recording four micrographs per ice hole. Movies were recorded at a magnification of 120,000×, corresponding to the 0.658 Å pixel size at the specimen level, with defocus values ranging from −1 to −2.4 µm. Exposures were obtained with 60 e⁻/Ų total dose. In total, 4,530 micrographs in EER format were collected.

Data processing was performed with cryoSPARC (version 4.4) including motion correction and contrast transfer function (CTF) determination. Particle picking and extraction (extraction box size 256 pixels) were carried out using cryoSPARC version 4.478. Next, several rounds of reference-free 2D classification were performed to remove artifacts and selected particles were used for ab initio reconstruction and heterorefinement. After heterorefinement, 18,497 particles contributed to an initial 3D reconstruction of 4.8 Å resolution (Fourier shell coefficient 0.143) with a C1 symmetry. A model of a Bet v1 (PDB ID 4A81) and AlphaFold2 (ColabFold implementation) models of the Fab08 and Fab10 were fitted into the cryo-EM maps with UCSF ChimeraX (version 1.5). These docked models were extended and rebuilt without lateral chain manually using Coot (version 0.9.8.8) and Phenix (version 1.21) (Emsley et al., 2010; Liebschner et al., 2019). Figures were prepared using ChimeraX (USCF) (Goddard et al., 2018).

NFAT-Luciferase Jurkat cells were purchased from BPS Bioscience Inc. (#60621) and cultured in RPMI 1640 (Gibco) supplemented with 10% FBS, 1% nonessential amino acids, 1% sodium pyruvate, 1% penicillin–streptomycin, and geneticin (1 mg/ml). Lentiviruses encoding the different anti-Bet v1 CAlleRs or the anti-CD19 CAR with mCherry reporter were used to transduce cells that were further sorted to purify mCherry⁺ cells. For K562 Bet v1 generation, the Bet v1 protein linked to the PDGF transmembrane domain and to GFP was cloned in the lentiviral expression vector pCDH-EF1-FHC plasmid (#64874; Addgene, RRID: Addgene_64874) (Yousefzadeh et al., 2014) under the EF1a promoter. Lentiviral particles were used to transduce tumor K562 cell line (RRID: CVCL_0063) that was further sorted to purify GFP⁺ cells. K562 cells were genetically modified to knock out CD32 (FcγRII) using CRISPR-Cas9. Briefly, 80 μM CRISPR RNA (crRNA) sequences targeting FCGR2A gene (5′-TGG​AGC​ACG​TTG​ATC​CAC​GG-3′ and 5′-AAA​GCA​CAG​TCA​GAT​GCA​CA-3′) (Synthego) and 80 μM trans-activating crRNA (Lot# 748524; IDT) were complexed with 45 μM high-fidelity recombinant Cas9 protein (produced by the Protein Production and Structure Core Facility of the EPFL—Swiss Federal Technology Institute of Lausanne, Switzerland) and mixed with 1 × 10⁶ K562 cells resuspended in 20 μl Lonza P3 Primary Cell Solution prior to electroporation in Lonza 4D-Nucleofector X Unit with the FF-120 program. After electroporation, cells were immediately transferred to prewarmed RPMI-1640 medium supplemented with 10% FBS and incubated at 37°C, 5% CO₂. CD32-deficient cells were further sorted. To generate K562 CD64 cells, CD32-deficient cells were transduced with a lentivirus encoding human CD64 and further sorted.

All cell lines were checked for Mycoplasma contamination with negative results.

Female BALB/cByJ mice aged 6 wk were purchased from Jackson Laboratory (RRID: IMSR_JAX:001026) and maintained at the animal facility of the University of Lausanne. All animal protocols were approved by the Cantonal Commission for Animal Experiments from the Canton of Vaud (Switzerland) (license number: VD3893).

Human T cells were enriched using EasySep Human T Cell Enrichment Kit following the manufacturer’s instructions (Cat. #17952; StemCell Technologies). For human Tregs, CD25⁺ PBMCs were enriched using CD25 MicroBeads II (Cat. #130-092-983; Miltenyi Biotec) and stained with CD4-FITC (BD), CD25(CD25-4E3)-APC, and CD127-PE for further sorting on a BD FACSAria II cell sorter (BD Biosciences) as CD4⁺CD25⁺CD127^low. Human primary cells were cultured for 2 days in X-Vivo 15 medium (Cat. #02-053Q; Lonza) supplemented with 5% human AB serum (Cat. #P30-2901; Pan-Biotech), 1% penicillin–streptomycin (Cat. #4-01F00-H; BioConcept), 55 μM 2-mercaptoethanol (Cat. #21985-023; Gibco), and 10 mM N-acetyl-L-cysteine (Cat. #A9165-25G; Sigma-Aldrich). After viral transduction, cells were cultured in RPMI medium (Cat. #61870-010; Gibco) supplemented with 10% FBS (Cat. #F7524; Sigma-Aldrich), 1% nonessential amino acids (Cat. #11140-050; Gibco), 10 mM HEPES (Cat. #15630-056; Gibco), 1 mM sodium pyruvate (Cat. #11360-039; Gibco), and 1% penicillin–streptomycin. Medium was supplemented with recombinant human IL-2 (Cat. #130-097-746; Miltenyi Biotec) at 30 IU/ml for Teff and 300 IU/ml for Tregs. Both T cells and Tregs were activated from day 0 with anti-CD3/CD28 Dynabeads (Cat. #11131D; Gibco) at 1:1 ratio.

Murine T cells were isolated from spleens and lymph nodes of female BALB/c mice. T cells were enriched by negative selection using EasySep Mouse T Cell Isolation Kit (Cat. #19851; StemCell Technologies). For Treg isolation, CD4⁺ T cells were isolated using EasySep CD4⁺ T Cell Isolation Kit according to the manufacturer’s instructions (Cat. #19852A; StemCell Technologies). CD4⁺ T cells were stained with CD25(PC61)-PE (Invitrogen) and CD4(RM4-5)-APC (BD), and the CD4⁺CD25^high fraction was purified using MoFlo Astrios (Beckman Coulter) or Aria II (BD) cell sorters. Murine T cells were cultured in supplemented RPMI cell culture medium with 55 μM 2-mercaptoethanol (Cat. #21985-023; Gibco). T cells were activated with mouse anti-CD3/CD28 Dynabeads (Cat. #11452D; Gibco) at 1:1 bead-to-cell ratio for conventional T cells and at 3:1 ratio for Tregs. The medium was supplemented with 50 IU/ml human IL-2 (Cat. #130-097-746; Miltenyi Biotec), 5 ng/ml human IL-7 (Cat. #130-095-362; Miltenyi Biotec), and human 5 ng/ml IL-15 (Cat. #130-095-764; Miltenyi Biotec) for conventional T cells. For Tregs, medium was supplemented with 2000 IU/ml recombinant human IL-2 (Cat. #130-097-746; Miltenyi Biotec) for Treg. On day 7 of culture, beads were removed and cells rested overnight before further experiments.

Second-generation CARs/CAlleRs with the CD28 hinge, CD28 transmembrane, CD28 intracellular domains, and the CD3ζ signaling domain were cloned in the pCDH-EF1-FHC lentiviral vector plasmid (no. 64874; Addgene) as previously described (Muller et al., 2021a). The CARs were designed using the scFv derived from the four novel anti-Bet v1 antibodies and the anti-CD19 scFv derived from clone FMC63 as previously described (Muller et al., 2021b).

The murine stem cell virus–based splice-gag vector (pMSGV) was used to introduce the scFv of either the CAlleR 8 or an anti-HLA A2 CAR (clone SN607D8) (Muller et al., 2021a; Wagner et al., 2022) fused to the hinge, transmembrane, and intracellular domains of mouse CD28, and the signaling domain of mouse CD3ζ. The retroviral vector was kindly provided by Dr. Melita Irving (Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland).

To produce lentiviruses, 3 × 10⁶ HEK293T cells were seeded in 9 ml of supplemented DMEM and transfected 24 h later with 4 μg of vector plasmid, 4 μg of pCMV-dR8.9 packaging plasmid, and 2 μg of pMD2.G2 packaging vector diluted in PEI. The supernatant was collected 48 and 72 h after transfection, filtered through a 45-μm filter, ultracentrifuged at 50,000g for 2 h, and stored at −80°C.

Retroviruses were produced using Platinum-E (Plat-E) Retroviral Packaging Cell Line (Cat. #RV-101; Cell Biolabs). Briefly, Plat-E cells were grown in DMEM supplemented with 10 μg/ml blasticidin (Cat. #ant-bl-1; InvivoGen) and 1 μg/ml puromycin (Cat. #P8833; Sigma-Aldrich) before transfection with 10 μg of CAR and 10 μg of pCL-Eco (#12371; Addgene plasmid, RRID:Addgene_12371) in Turbofect reagent (Cat. #R0531; Thermo Fisher Scientific). Two days after transfection, Plat-E supernatant was collected, and viruses were concentrated 400 times after ultracentrifugation at 50,000g for 1 h 30 min at 4°C.

For human cells, transduction was performed 24 h after activation by adding a MOI of 1 of pretitrated lentivirus to the culture. After 20 h of incubation, the virus was washed, and cells were resuspended in fresh supplemented medium with cytokines. For murine cells, transduction was performed 24 h after activation for conventional T cells and 72 h after activation for Tregs. Cells were incubated with 10% concentrated retrovirus in a nontreated culture plate (Greiner Bio-One) precoated overnight with 20 µg/ml RetroNectin (Cat. #T100A; Takara Bio). The cell–virus mixture was spinoculated at 2,000g for 1.5 h at 32°C, followed by overnight incubation at 37°C, 5% CO₂. After 18 h of transduction, the virus was removed by washing, and cells were resuspended in fresh medium.

To screen CAlleR functionality, NFAT cells were cocultured with K562 WT or K562 expressing Bet v1 at 2:1 ratio for 24 h at 37°C, 5% CO₂. After incubation, the luciferase assay was performed as previously described²³. Briefly, cells were lysed for 20 min under shaking conditions with 50 μl harvesting buffer (50 mM 2-morpholinoethanesulfonic acid sodium [NaMES] [Cat. #1061970100; Sigma-Aldrich], 50 mM Tris-HCl, 1 mM dithiothreitol [Cat. #R0861; Thermo Fisher Scientific], 0.4% Triton X-100 [Cat. #9002-93-1; Thermo Fisher Scientific]). After incubation, 50 μl luciferase buffer (125 mM NaMES, 125 mM Tris-HCl, 25 mM [CH₃COO]₂Mg · 4H₂O [Cat. #M0631; Sigma-Aldrich], 2.5 mM adenosine triphosphate [ATP, Cat. #R0441; Thermo Fisher Scientific]) was added and incubated for 1 min under shaking conditions prior to adding 50 μl of luciferin buffer (1 mM D-luciferin [Cat. #88292; Thermo Fisher Scientific] in 4.8 mM KH₂PO₄ [Cat. #1.04873; Merck] and incubating for 5 min protected from light. Luminescence was measured as relative light units on the Synergy H1 Hybrid reader [BioTek]).

For activation and proliferation assays involving human primary T cells and Tregs, 0.1 × 10⁶ primary cells were cocultured with 120 G irradiated K562 Bet v1 at different ratios, or when not specified at 1:5 ratio. For activation assay testing different FcγRs, 120 G irradiated K562 at 1:5 ratio was used in the presence of the Bet v1 protein (1 μg/ml), anti-Bet v1 mAbs (0.5 μg/ml), and/or decomplemented serum from birch-allergic donors. After 2 days of coculture, cells were stained for activation markers and analyzed by flow cytometry. Anti-CD3 was used as a positive control (1 μg/ml, clone OKT3, Cat. #566685; BD Biosciences). For activation of CAlleR T cells cocultured with other CAlleR T cells, 1:1 ratio was used in the presence of Bet v1 (0.5 μg/ml).

Each replicate of activation and suppression assays corresponded to an independent donor.

In vitro activation of CAlleR murine cells was assessed by coculturing 0.1 × 10⁶ T cells or Tregs with 0.02 × 10⁶ BMDCs generated as previously described (Alcaraz-Serna et al., 2021) in the presence of Bet v1 (0.5 μg/ml) and 10% heat-inactivated murine serum. Cells were incubated for 24 h, and upregulation of activation markers was assessed by flow cytometry.

BPE (European white Betula pendula, Cat. #XP527D3A25; Stallergenes Greer) was reconstituted in sterile PBS at 3.33 mg/ml. On days 0, 2, 4, 7, 9, and 11, mice were anesthetized by inhalation in a chamber with 4% isoflurane for 1 min and intranasally instilled with 30 μl of diluted BPE. On day 0 (prophylactic model) or on day 8 (treatment model), mice were randomly assigned to receive adoptive transfer of CAR or CAlleR murine Tregs or vehicle (PBS). On days 14 or 15, mice were euthanized with 225 mg/kg intraperitoneal injection of pentobarbital (Esconarkon, Streuli Pharma AG).

After sacrifice, blood was collected from abdominal aorta for serum isolation. Lungs were extracted, cut, and digested for 30 min at 37°C with 0.2 mg/ml of collagenase D (Cat. #11088858001; Sigma-Aldrich) resuspended in HBSS (Gibco), 10 mM HEPES, and 5% FBS. mLNs were collected and digested the same way as lungs. Organs were then smashed on a 70-μm cell strainer with syringe plungers and stained for flow cytometry analysis. Spleens were directly smashed on 70-μm cell strainers, and erythrocytes were lysed with ACK lysis buffer (155 mM NH₄Cl, 10 mM KHCO₃, 0.1 mM EDTA in distilled water, pH 7.4) prior to staining for flow cytometry.

For BALF collection, tracheas were cannulated with an 18 G Venflon (Cat. #393226; BD) and flushed three times with 500 μl of ice-cold PBS supplemented with 0.2% BSA (Cat. #A8806; Sigma-Aldrich). Total BALF cell counts were determined with Coulter Counter Multisizer 4e (Cat. #B23005; Beckman Coulter). Cytospins were obtained by centrifuging 80,000 cells at 800 rpm for 5 min. Resulting cytospin slides were stained with RAL Diff-Quik solution (Cat. #720555-0000; CellaVision). Percentages of monocytes/macrophages, eosinophils, neutrophils, and lymphocytes were assessed by counting 200 cells per cytospin.

For human samples, cells were washed once with PBS supplemented with 0.5% FBS, 0.4% EDTA. After discarding supernatant, cells were incubated for 30 min at 4°C in the same wash buffer with diluted antibodies: anti-human CD3 (UCHT-1), PECy7 (1:400 dilution), 563423; BD Biosciences, RRID:AB_2738196; anti-human CD4 (RPA-T4), AF700 (1:400 dilution), 557922; BD Biosciences; anti-human CD8 (RPA-T8), APC-Cy7 (1:400 dilution), 557760; BD Biosciences; anti-human CD8 (SK1), PE (1:250 dilution), 345773; BD Biosciences; anti-human CD25 (M-A251), FITC (1:400 dilution), 555431; BD Biosciences; anti-human CD25 (CD25-4E3), APC (1:250 dilution), 17-0257-42; Thermo Fisher Scientific; anti-human CD64 (REA978), APC (1:200 dilution), 130-116-197; Miltenyi Biotech; anti-human CD71 (M-A712), FITC (1:400 dilution), 555536; BD Pharmingen; anti-human CD127 (hIL-7R-M21), PE (1:200 dilution), 557938; BD Biosciences; anti-human CD32 (FLI8.26), APC (1:200 dilution), 559769; BD Biosciences; anti-human CD16 (3G8), PE (1:200 dilution), 555407; BD Biosciences; biotinylated rBet v1 (1 μg/ml); streptavidin, PE (1:1,000 dilution), 554061; BD Biosciences; DAPI (Live/Dead), Invitrogen (1:2,500 dilution), D1306; Phantom Dye (Live/Dead), Pacific Blue (1:2,500 dilution), PD00004; Proteintech. After 30-min incubation, samples were washed again and either resuspended in washing buffer prior to flow cytometry acquisition, or permeabilized and fixed using eBioscience Foxp3/Transcription Factor Staining Buffer according to the manufacturer’s instructions (Cat. #00-5523-00; Invitrogen). Intracellular staining was performed with anti-human Foxp3 (PCH101), eFluor660 (1:250 dilution), 50-4776-42; Thermo Fisher Scientific; Helios (22F6), PE (1:250 dilution), 137216; BioLegend.

For murine samples, cells were incubated with an anti-mouse CD16/32 antibody (TruStain FcX PLUS, clone S17011E; BioLegend (1:250 dilution), Cat. #156604, RRID:AB_2783138) for 15 min at 4°C. After blocking, cells were washed and extracellular staining was performed for 30 min at 4°C with the following antibodies: anti-mouse CD3 (145-2C11), PE-Cy7 (1:250 dilution) and APC-Cy7 (1:100 dilution), 552774; BD Biosciences (RRID:AB_394460) and 561042 (RRID:AB_2034003), respectively; anti-mouse CD4 (RM4-5), PerCP-Cy5.5 (1:300 dilution) and APC (1:250 dilution), 561090 and 553051; BD Biosciences, respectively; anti-mouse CD8a (53-6.7), FITC (1:100 dilution) and Pacific Blue (1:400 dilution), 553030 and 558106; BD Biosciences, respectively; anti-mouse CD25 (PC61.5), PE (1:200 dilution), 12-0251-83; Invitrogen; anti-mouse CD25 (PC61), PerCP-Cy5.5 (1:250 dilution), 551071; BD Biosciences; anti-mouse CD44 (IM7), FITC (1:100 dilution), 553133; BD Biosciences; anti-mouse CD45 (30-F11), PE-Cy7 (1:250), 552848; BD Biosciences; anti-mouse CD45.2 (104), AF700 (1:100 dilution), 109821; BioLegend; anti-mouse CD69 (H1.2F3), PE (1:400 dilution), 553237; BD Biosciences; anti-mouse CD86 (GL1), PE (1:250 dilution), 553692; BD Biosciences; anti-mouse CD90.1 (Thy1.1; OX-7), PerCP-Cy5.5 (1:250 dilution), 557266; BD Biosciences; anti-mouse CD90.1 (Thy1.1; REA838), FITC (1:500 dilution), 130-112-872; Miltenyi Biotech; anti-mouse CD11c (HL3), APC (1:100 dilution), 550261; BD Biosciences; anti-mouse CD11c (N418), APC-Cy7 (1:600 dilution), 117323; BioLegend; anti-mouse CD11b (M1/70), PECy7 (1:600 dilution), 561098; BD Biosciences; anti-mouse MHC-II (M5/114.15.2), AF700 (1:1,000 dilution), 107621; BioLegend; anti-mouse F480 (BM8), PerCP (1:100 dilution), 123126; BioLegend; anti-mouse Ly6C (HK1.4), Pacific Blue (1:1,000 dilution), 128013; BioLegend; anti-mouse SiglecF (E50-2440), PE (1:100 dilution), 562068; BD Biosciences; anti-mouse Ly6G (1A8), PerCP Cy5.5 (1:100 dilution), 560602; BD Biosciences; biotinylated rBet v1 (1 μg/ml); streptavidin, APC (1:500 dilution) and PE (1:1,000 dilution), 554067 and 554061; BD Biosciences, respectively; Aqua (Live/Dead), AmCyan (1:1,000 dilution), L34957; Invitrogen; DAPI (Live/Dead), Invitrogen (1:2,500 dilution), D1306. After incubation, samples were washed again and either fixed with BD FACS Lysing Solution following the manufacturer’s instructions (Cat. #349202; BD Biosciences) or permeabilized and fixed using eBioscience Foxp3/Transcription Factor Staining Buffer set (Cat. #00-5523-00; Invitrogen) prior to intracellular staining. Intracellular staining was performed with anti-mouse Foxp3 (FJK-16 s), FITC (1:250 dilution) and PE (1:250 dilution), 11-5773-82 and 12-5773-82; Invitrogen, respectively; anti-mouse IL-4 (BVD6-24G2), PECy7 (1:100 dilution), 25-7042-41; Invitrogen; anti-mouse IL-5 (TRFK5), APC (1:100 dilution), 562048; BD Biosciences; anti-mouse IL-13 (eBio13A), ef450 (1:100 dilution), 48-7133-80; Invitrogen; anti-mouse IL-17 (eBio17B7), PE (1:100 dilution), 12-7177-81; Invitrogen.

Single-cell suspensions were resuspended in supplemented PBS prior to acquisition by an LSRFortessa cell analyzer (BD Biosciences). Flow cytometry data were analyzed using FlowJo software version 10.9.0.

Mice were anesthetized with intraperitoneal injection of 50 mg/kg of pentobarbital and intramuscular injection of 100 mg/kg ketamine. Mice were tracheotomized, and a 18 G metallic cannula was inserted in the trachea and secured with sutures. Mice were connected to the FlexiVent FX (SCIREQ Scientific Respiratory Equipment Inc.) and mechanically ventilated at 150 breaths/min, a tidal volume of 10 ml/kg, and a PEEP set at 3 cm H₂O. Forced oscillation perturbation was performed to test the lung to a standardized signal of oscillatory frequencies above and below the normal ventilation frequency. Changes in resistance to increasing concentrations of nebulized acetyl-b-methacholine chloride (Sigma-Aldrich) were measured from snapshot perturbation measurements taken using the forced oscillation perturbation technique.

Mice were euthanized, and lungs, mLNs, and spleen were dissected and fixed in 4% paraformaldehyde (PFA, Cat. #158127; Sigma-Aldrich) overnight at 4°C under shaking conditions. Organs were then washed three times in PBS and incubated in PBS with 30% sucrose overnight at 4°C prior to embedding in Tissue-Tek (Cat. #4583; Sakura) and snap-frozen in dry ice. Organ sections were cut with 10 μm thickness and assembled in microscopy slides for further staining. Slides were fixed in 4% PFA for 5 min, washed three times with room-temperature PBS, and incubated at RT in blocking buffer (PBS supplemented with 5% donkey serum, 0.5% BSA, 0.1% Triton X-100, 0.01% sodium azide) for 30 min. Primary antibodies (rabbit anti-GFP [Cat. #ab290; Abcam, RRID:AB_303395; 1:2,000 dilution], rat anti-mouse B220 [Cat. #14-0452-81; Thermo Fisher Scientific, RRID:AB_467253; 1:600 dilution], armenian hamster anti-mouse CD11c [Cat. #550283; BD Biosciences; 1:200 dilution]) were diluted in blocking buffer and incubated overnight at 4°C. Slides were then washed with PBS/0.3% Triton X-100, and secondary antibodies (IgG donkey anti-rabbit-AF488 [Cat. #A-21206; Thermo Fisher Scientific, RRID:AB_2535792; 1:300 dilution], IgG chicken anti-rat-AF647 [Cat. #A-21472; Thermo Fisher Scientific, RRID:AB_2535875; 1:300 dilution], IgG goat anti-armenian hamster-biotinylated [Cat. #127-065-160; Jackson Immuno Research, RRID:AB_2338980; 1:600 dilution]) were diluted in blocking buffer and incubated for 1 h at RT. After secondary staining, washing step was repeated three additional times and streptavidin-AF555 (Cat. #S32355; Thermo Fisher Scientific; 1:300 dilution) was diluted in blocking buffer, added to the slides, and incubated for 1 h at RT. Slides were washed again and mounted with Fluoromount-G Mounting Medium with DAPI (Cat. #00-4959-52; Thermo Fisher Scientific) prior to storage at 4°C.

Images were acquired at 20× magnification with a Hamamatsu NanoZoomer S60 slide scanner and analyzed with QuPath software version 0.5.1 (Bankhead et al., 2017). Briefly, whole-slide images were imported into the software, and after necessary adjustments to color normalization, regions of interest (ROIs) were manually annotated to analyze whole stained tissues. A machine learning–based classifier was developed in QuPath to differentiate positive and negative cells for each channel. Classifier’s performance was improved by refining the training dataset and adjusting classifier parameters. Then, automated cell detection was performed using QuPath’s built-in cell detection algorithm. The trained classifier was then applied to the detected cells, categorizing them as positive or negative based on the learned features. The number of GFP⁺ cells was calculated for each ROI, and density of cells was assessed by dividing the number of detected positive cells by the area of ROI. Cell interactions were quantified by determining the number of double-positive cells.

For experiments involving human primary cells, each replicate corresponded to a unique healthy donor. In animal experiments, each replicate represented a single mouse, except for immunofluorescence staining, where two separate slides per organ were independently prepared, stained, and analyzed as technical replicates. Statistical analyses were performed using GraphPad Prism version 10.4.1 (GraphPad Software). The specific statistical tests used for each experiment are detailed in the corresponding figure legends, along with exact P values. Normality of the data was assessed using the Shapiro–Wilk test.

Fig. S1 supports Fig. 1 by providing human B cell sorting strategy for antibody discovery and characterization of three novel human anti-Bet v1 IgE antibodies. Fig. S2 supports Fig. 4 by showing the generation and validation of K562 cell lines expressing different FcγRs. Fig. S3 supports Fig. 4 by demonstrating CAlleR T cell activation by soluble Bet v1 through noncompetitive scFvs. Fig. S4 supports Fig. 6 by showing additional experiments performed in the in vivo treatment model. Fig. S5 supports Fig. 7 by showing additional experiments from the in vivo preventive model.

Data shown in the article and supplementary information are available upon reasonable request. The accession numbers for the sequences of all novel anti-Bet v1 antibodies reported in this study are GenBank: PZ143166–PZ143179. Cryo-EM map is available on Zenodo (https:/doi.org/10.5281/zenodo.16575186) and will be made public upon publication.

We thank the Immunology and Allergy Division of the Centre Hospitalier Universitaire Vaudois and the Center for Immunotherapy and Vaccinology for their support. We are grateful to the Dubochet Center for Imaging (Lausanne) at the École Polytechnique Fédérale de Lausanne (EPFL) and University of Lausanne (Unil) for their assistance and to the Protein Production and Structure Core Facility (PTPSP) at EPFL for their help in the antibody production and characterization. We thank the Dubochet Center for Imaging (an EPFL, UNIGE, UNIL initiative) for cryo-EM grid preparation and data collection. We also thank the Cellular Imaging Facility of Unil for assistance with immunofluorescence image analysis, and the Unil’s animal facility for their dedicated animal care. Finally, we thank the patient and healthy donors who voluntarily contributed to this study.

Y.D. Muller received funding for this project from the Gabriella Giorgi-Cavaglieri Foundation, and A. Alcaraz-Serna was supported by educational grants from the Martin Escudero Foundation and the Machaon Foundation.

Author contributions: Ana Alcaraz-Serna: formal analysis, funding acquisition, investigation, methodology, project administration, validation, visualization, and writing—original draft, review, and editing. Noémie Chillier: formal analysis, investigation, and writing—review and editing. Aurélien Trompette: conceptualization, investigation, methodology, and writing—review and editing. Laura Ermellino: investigation, methodology, and writing—review and editing. Raphaël Porret: investigation, methodology, and writing—review and editing. Erica Lana: investigation and writing—review and editing. Rebecca Cecchin: investigation and writing—review and editing. Apiha Shanmuganathan: investigation, resources, and writing—review and editing. Seher Guney: investigation and resources. Oscar Alfageme-Abello: investigation and writing—review and editing. Eleonora Pace: investigation, resources, and writing—review and editing. Antonio Mancarella: resources. Jeremy Campos: formal analysis. Mathilde Foglierini: formal analysis. Christophe von Garnier: funding acquisition, resources, supervision, and writing—original draft, review, and editing. Craig Fenwick: data curation, formal analysis, methodology, and supervision. Laurent Perez: formal analysis, supervision, and writing—review and editing. Niki Ubags: investigation, methodology, resources, validation, and writing—review and editing. Yannick D. Muller: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.