Defatted pollen of P. pratense and B. verrucosa (Iber-Polen, Jaén, Spain) were extracted as previously described for P. pratense (22). Briefly, pollen was stirred in phosphate-buffered saline (PBS; pH 7.4, 4°C), clarified by centrifugation (15,000 × g, 10 min, 4°C), subjected to tangential flow ultrafiltration (cutoff pore size, 100 kDa), dialyzed against water, sterile-filtered (0.22 µm), and lyophilized. Freeze-dried aliquots were stored at −80°C and served as NEs (Pp-N and Bv-N for P. pratense and B. verrucosa NEs, respectively).

NEs were reconstituted at 2 mg protein/ml in 0.1 M phosphate buffer (pH 7.4). The protein content prior to polymerization was determined with the Bradford assay (Bio-Rad, Hercules, CA, USA). Cross-linking reactions were performed at a final volume of 20 ml by adding GA and PBS to reach five protein-to-GA ratios, i.e., 0.1×, 0.2×, 1×, 5×, and 10×, relative to the standard reference condition of 25 mM GA (designated as P1). The resulting preparations—P0.1 (2.5 mM GA), P0.2 (5 mM GA), P1 (25 mM GA), P5 (125 mM GA), and P10 (250 mM GA)—were incubated at 22°C for 3 h under stirring. The reactions were quenched with glycine using 125 mM for the standard reference condition (P1, 25 mM GA) and adjusting the glycine concentration proportionally for all other GA concentrations. Samples were sterile-filtered (0.22 µm), and excess of GA, non-polymerized allergens, and other residual substances were removed with five cycles of ultrafiltration using Vivaspin 20 protein concentrator devices (100 kDa cutoff; Sartorius, Göttingen, Germany) through buffer exchange from PBS to Milli-Q water to guarantee conductivity values below 1,000 µS/cm. The flow-through obtained after the first ultrafiltration step, corresponding to the low-molecular-weight fraction (LMWF), was retained for analysis, while the retained high-molecular-weight fraction (HMWF) was considered the PEs. The protein content of both fractions was determined with the Bradford assay. The polymerization efficiency (Table 1) was estimated from the protein concentration in the LMWFs. As the initial protein content in the extracts is known, the fraction of protein remaining in the LMWFs was used to infer the amount retained in the HMWF (PEs).

PEs were sterile-filtered, lyophilized, and stored at −80°C until used. The final Pp-N- and Bv-N-derived samples were designated as Pp-P10, Pp-P5, Pp-P1, Pp-P0.2, and Pp-P0.1 and as Bv-P10, Bv-P5, Bv-P1, Bv-P0.2, and Bv-P0.1, respectively.

Native Bet v1 and Phl p5 were purified as previously described (23) and served as monospecific controls.

Of the NEs or PEs, 5 μg was mixed with loading buffer and boiled for 5 min. The sample was separated in Any kD™ Mini-PROTEAN^® TGX™ Precast Protein Gelsat (Bio-Rad) or 4%–20% Mini-PROTEAN^® TGX™ Precast Protein Gels, at 280 V, using a Mini-PROTEAN II apparatus (Bio-Rad). Precision Plus Protein All Blue Standards (Bio-Rad) was used as the molecular weight (MW) reference. Visualization of the separated molecules was achieved by staining with colloidal Coomassie (GelCode Blue Stain Reagent; Life Technologies, Carlsbad, CA, USA). Gels were documented using a ChemiDoc Imaging System (Bio-Rad) with the “Protein gels–Coomassie stain” preset and subsequent color rendering in Image Lab 6.1 software to show the Coomassie blue staining.

For Western blot (WB), proteins in the polyacrylamide gel were transferred onto a 0.45-µm nitrocellulose membrane using a Transblot Semi-Dry Electrophoresis transfer cell (Bio-Rad) at 20 V for 30 min. The membranes were blocked with PBS containing 0.25% Tween-20 (PBS-T) containing 5% bovine serum albumin (BSA) for 60 min. The membranes were incubated with sera from individuals sensitized to P. pratense or B. verrucosa or 1:5,000 diluted sera from mice immunized with Phl p 5, Bet v 1, Pp-N, or Bv-N. For the IgE binding studies, five pooled sera (1:50) from subjects sensitized to Pp-N (AbBaltis, Sittingbourne, UK) and Bv-N (PlasmaLab, Everett, WA, USA) were used. For murine IgG reactivity, the membranes were incubated with mixtures of two sera from mice immunized with Phl p 5, Bet v 1, Pp-N, or Bv-N diluted 1:10,000. After incubation, the membranes were incubated with 1:2,000 horseradish peroxidase (HRP)-conjugated mouse anti-human IgE (SouthernBiotech, Birmingham, AL, USA) or 1:2,000 HRP-labeled anti-mouse IgG (Sigma-Aldrich, St. Louis, MO, USA). Protein–antibody binding was detected with an ECL Prime detection reagent (GE Healthcare Life Sciences, Little Chalfont, UK) in a ChemiDoc imaging system (Bio-Rad). Precision Plus Protein WesternC Standards (Bio-Rad) was used as the MW marker, previously incubated with Precision Protein StrepTactin–HRP Conjugate (Bio-Rad) for detection.

The protein and allergen composition were determined by MS (24), conducted at the Proteomics Unit, National Center for Biotechnology (CNB-CSIC, Madrid, Spain). Samples were resuspended in 500–1,000 µl of 2.5% (w/v) SDS, 25 mM triethylammonium bicarbonate (TEAB), 5 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), and 10 mM chloroacetamide (CAA). Lysates were incubated at 60°C for 30 min to reduce and alkylate protein cysteine residues. The protein extracts were centrifuged at 18,000 × g for 10 min, and the supernatant was transferred into a new tube and quantified using Pierce™ 660 nm Protein Assay Reagent supplemented with Ionic Detergent Compatibility Reagent according to the manufacturer’s instructions. For automatic SP3-based protein digestion, the samples were processed with the Opentrons OT-2 robot in the presence of MagReSyn^® HILIC microparticles (ReSyn Biosciences, Edenvale, South Africa) and digested at 37°C overnight using trypsin (1:33 enzyme/protein ratio) and Lys-C (1:500 enzyme/protein ratio) enzymes. The eluted peptides were dried in a speed vacuum and quantified by fluorimetry (QuBit, Paris, France) according to the manufacturer’s instructions.

For nano-liquid chromatography coupled to electrospray ionization tandem mass spectrometry (nanoLC-ESI-MS/MS) analysis, 500 ng of each sample was individually analyzed using an UltiMate 3000 Nano HPLC System (Thermo Fisher Scientific, Waltham, MA, USA) coupled online to an Orbitrap Exploris™ 240 mass spectrometer (Thermo Fisher Scientific). Each sample [500 ng in 5 µl of sample resuspended in mobile phase A: 0.1% formic acid (FA)] was loaded on a 50-cm × 75-μm µPAC Neo C18 analytical column at 50°C and separated at a flow rate of 250 nl/min using a 120-min gradient ranging from 2% to 95% mobile phase B [80% acetonitrile (ACN) in 0.1% FA]. Data acquisition was performed using a data-dependent top 20 method, in full scan positive mode, scanning 375–1,200 m/z. MS1 scans were acquired at an Orbitrap resolution of 60,000 at m/z 200, with normalized automatic gain control (AGC) target of 300%, a radio frequency (RF) lens of 80%, and an automatic maximum injection time (IT). The 20 most intense ions from each MS1 scan were selected and fragmented with a higher-energy collisional dissociation (HCD) of 34%. The resolution for the HCD spectra was set to 45,000 at m/z 200, with a normalized AGC target of 50% and an automatic maximum IT. Isolation of precursors was performed with an isolation window of 0.7 m/z and 45 s of exclusion duration. Precursor ions with charge states from 2⁺ to 5⁺ were included.

Raw instrument files were processed using Proteome Discoverer (PD) version 3.1.1.93 (Thermo Fisher Scientific). MS2 spectra were searched using Mascot Server v2.8.1 (Matrix Science, London, UK) against the Poeae or Betulaceae UniProtKB databases containing the most common laboratory contaminants (cRAP database with 70 sequences). All searches were configured with dynamic modifications for pyrrolidone from Q (−17.027 Da) and oxidation of methionine residues (+15.9949 Da) and static modification as carbamidomethyl (+57.021 Da) on cysteine, monoisotopic masses, and trypsin cleavage (maximum of two missed cleavages). The peptide precursor mass tolerance was 10 ppm, and the MS/MS tolerance was 0.02 Da. The false discovery rate (FDR) for proteins, peptides, and peptide spectral matches (PSMs) was kept at 1%.

Precursor ion quantitation was also performed in Proteome Discoverer using the “Minora” feature in the processing method and the “Feature Mapper” and “Precursor Ions Quantifier” nodes in the consensus step. Protein abundance was calculated by summing the sample abundance rates of the connected peptide groups (using unique+razor peptides).

One-dimensional ¹H spectra and a pseudo-two-dimensional (2D) diffusion-ordered spectroscopy (DOSY) were acquired in a Bruker Avance NEO 400-MHz spectrometer using standard pulse sequences included in TopSpin 4.1.1 software to analyze the physicochemical properties, the molecular size, and the stability of the native and polymerized preparations following standard protocols from our laboratories.

Briefly, the ¹H spectra were recorded by applying a zg30 pulse sequence, which presents a 30° excitation pulse instead of the conventional zg program (90° excitation pulse). Samples were prepared in D₂O, and experiments were acquired at 298 K on a 400-MHz Bruker Avance NEO spectrometer. Data were processed with MestReNova 15.0.1 (Mestrelab, Santiago de Compostela, Spain) and TopSpin 4.1.1 (Bruker, Madrid, Spain) software. These experiments were acquired with 64 scans, a spectral width (SW) of 14 ppm, and a free induction decay (FID) size of 65k data points. The DOSY spectra were obtained using the standard pulse sequence, acquiring 32 gradient points, with 16 scans each, between 2% and 98% gradient intensity using a diffusion time delay of big delta (Δ) 0.30 s and little delta (δ) 2,000 μs to achieve a wide pulse gradient. Data were analyzed using the pseudo-2D spectra and also obtaining the sum projection in the diffusion dimension of each one. Successful polymerization was inferred from line broadening, chemical shift dispersion, and decreased diffusion coefficients. Extract stability was assessed over 3 months at 4°C by monitoring changes in the integrated signal intensities relative to the baseline spectrum (25).

For the estimation of MWs, a calibration protocol using DOSY experiments was performed as previously described (26), with protein standards at 5 mg/ml in D₂O: ovalbumin (44 kDa), conalbumin (75 kDa), thyroglobulin (669 kDa), and Blue Dextran (2,000 kDa), all from Bio-Rad. Proteins were registered following the pulse sequence ledpbgp2s (Bruker). Due to the heterogeneous nature of the extract, eight signals with varying diffusion coefficient values were selected for calculation to reach a closer approximation of the molecular size distribution for each sample. The corresponding logD values were extrapolated from the calibration curve and were used to estimate the MW of each component in the preparation.

Experiments were carried out at the Spanish National Centre for Electron Microscopy (ICTS) at the Complutense University, Madrid, Spain. Samples (1,000 µg protein/ml) were adsorbed onto carboncoated copper grids, negatively stained with 2 % (w/v) uranyl acetate, and examined on a JEOL JEM-1010 TEM operated at 80 kV and fitted with a Gatan Erlangshen ES500W CCD camera (Gatan, Pleasanton, CA, USA). Apparent porosity was estimated from micrographs using ImageJ by segmenting the particle areas and calculating the mean gray value distributions. Darker, highdensity regions were interpreted as thicker, less porous domains.

Experiments were carried out at the Molecular Interactions Facility at the CIB Margarita Salas. Measurements were performed using a Protein Solutions DynaPro MS/X instrument (Protein Solutions, Ankeny, IA, USA) at 20°C and a 90° light scattering cuvette. Data were collected and analyzed with the Dynamics v6 software. Particle size distributions are reported as intensity-weighted hydrodynamic radii.

We thank the staff of CIB Margarita Salas Molecular Interactions Facility for technical assistance with the dynamic light scattering (DLS) assays.

Detection of human specific IgE against native or PEs was performed with ELISA using the five commercial sera mentioned above.

All sera were tested in titration experiments to determine the dilution to reach optical densities (ODs) of ~1.5 when tested against the NEs. The NEs and PEs were tested at the defined dilution. Briefly, microplates were coated overnight at 4°C with 10 μg/ml of the antigens in 0.05 M carbonate–bicarbonate buffer, pH 9.4. After three washes with PBS-T, nonspecific reactivity was blocked by 1 h of incubation with PBS-T containing 1% BSA. Wells were washed again and incubated overnight with diluted sera at 4°C. Subsequently, the wells were washed again and incubated for 1 h with 1:2,000 HRP-conjugated mouse anti-human IgE (SouthernBiotech). After a final washing step, 3 mg/ml of the peroxidase substrate o-phenylenediamine dihydrochloride (OPD) (Sigma-Aldrich) dissolved in citrate buffer (0.1 M citrate buffer with 0.03% H₂O₂, pH 5.5) was added to the wells to allow detection of the reaction. The reaction was stopped by adding 10% HCl and read at 492 nm in a spectrophotometer (Synergy Mx, BioTek, Winooski, VT, USA).

Competitive ELISA was performed to analyze the capacity of the preparations to block the IgE reactivity against the NEs and expressed as biological potency (BP; micrograms of competitor necessary to block 50% of the reaction). ELISA experiments were performed as described; however, serial dilutions of the inhibitor (NE or PE; 0.2–0.0004 µg/ml) were pre-incubated with pooled serum (1:60) for 2 h at 37°C and transferred to the coated wells. After completion of all steps, the inhibition percentage was calculated using the following equation: percent inhibition = 100 − [(OD of serum with competitor/OD of serum without competitor) × 100]. Sera adsorbed with PBS-T containing 1% BSA without competitor served as a positive control.

To measure the strength of the binding between PEs and the IgE, a potassium thiocyanate (KSCN) ELISA-elution assay was performed (27). The experimental procedure was similar to the ELISA for the IgE binding determination, with three differences: 1) the serum pool was added to the microplates at a dilution to reach ODs of ~1.5; 2) the incubation with sera lasted 2 h; and 3) after incubation with sera, KSCN diluted in PBS-T containing 1% BSA was added to the plates for 30 min at room temperature. The KSCN molar concentrations were: 2.8, 2.5, 2.2, 1.9, 1.6, 1.3, 1, 0.7, 0.4, and 0.1. The reaction was revealed as described above.

The results were expressed as percentage of reduction and avidity index, defined as the KSCN concentration required to elute 50% of sample–IgE complexes. To determine the avidity index, we calculated the reduction of absorbance after KSCN treatment relative to the samples without treatment.

Tests were performed using the mast cell line LAD2, kindly provided by Drs. Metcalfe and Kirshenbaum (National Institute of Allergy and Infectious Diseases, Bethesda, MD, USA) (28). LAD2 cells were cultured in complete StemPro-34 media (Gibco, Waltham, MA, USA) at 37°C and 5% of CO₂, as described (29). In addition, to increase the expression of the high-affinity IgE receptor (FcϵRI) before any experiment, the cells were cultured for 5 days with 10 ng/ml of recombinant human interleukin-4 (IL-4; R&D Systems, Minneapolis, MN, USA) before any experiments.

LAD2 cells were sensitized as reported (29). In brief, 10⁶ cells/ml were incubated overnight with a pool of sera from five patients allergic to B. verrucosa or P. pratense (1:25). Subsequently, the cells were centrifuged at 265 × g for 5 min and resuspended in fresh media. LAD2 cells were then challenged for 10 min in a 96-well flat-bottomed plate (0.5 × 10⁶ cells/ml) with increasing doses (0.1, 0.5, 1, 5, 25, and 50 μg/ml) of native or PEs of B. verrucosa and P. pratense. On the other hand, for the analysis of individual patients, sensitization was performed with serum from each subject (1:25) and the cells were activated with both extracts at doses of 5 and 10 μg/ml. After stimulation, flow cytometry analysis was performed as described (30). The cells were blocked with human TruStain FcX (1:50; BioLegend, San Diego, CA, USA) in ice-cold FACS buffer (2.5 mM EDTA and 0.5% BSA in PBS) for 15 min. The cells were stained for 30 min with a mix of fluorochrome-conjugated antibodies/FITC–FcεRI (AER-37; 1:100), PE–CD117 (104D2; 1:600), BV421–CD63 (H5C6; 1:200), APC–CD107a (H4A3; 1:100), and PE–cyanine7–IgE (MHE-18; 1:100), all from BioLegend. Moreover, cell viability was assessed with eFluor780 dye (1:2,000; Thermo Fisher Scientific). Finally, the cells were washed with FACS buffer and analyzed in a FACSCanto II flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA). On average, 10,000 events of live and single cells were recorded. Analysis was performed with the FlowJo v10 program. Non-stimulated and polyclonal α-human IgE-stimulated cells (Sigma-Aldrich) were used as negative and positive activation controls, respectively. A representative gating strategy for a positive sample is shown in Supplementary Figure 1, outlining the sequential gating steps applied for the flow cytometry analysis.

The activity of β-hexosaminidase was measured as previously reported, with minor modifications (29). In brief, 40 µl of the supernatant from each condition was placed in a 96-well flat-bottomed plate, in duplicate, and 40 µl of the β-hexosaminidase substrate solution (2 mM p-nitrophenyl N-acetyl β-d-glucosamine diluted in 100 mM citrate buffer) was added. The culture medium alone and the supernatants from Triton-lysed cells were included to determine the absorbance background and to measure the total levels of β-hexosaminidase, respectively. Subsequently, the plate was incubated for 45 min at 37°C in the dark and 80 µl of 1 M NaOH was added afterward to stop the reaction. Finally, the absorbance (OD) was measured at 405 nm using the GloMax Discover (Promega, Madison, WI, USA) microplate reader. The percentage of β-hexosaminidase activity was calculated according to the following formula:

Degranulation (%) = [(OD supernatant − OD background)/(OD total lysis − OD background)] × 100

BALB/cByJ mice were obtained from Charles River Laboratories (France). All animal experiments were conducted at the Experimental Medicine and Surgery Department, Hospital Clínico San Carlos (Madrid, Spain), in accordance with the ARRIVE guidelines and following the European Directive 2010/63/EU on the protection of animals used for scientific purposes. National regulations were also followed, specifically Spanish R.D. 53/2013. All procedures were approved by the Institutional Animal Care and Ethics Committee.

Immunization of mice was performed to examine the immunogenicity and the humoral responses induced by the Pp-P1 and Bv-P1 preparations and Pp-N and Bv-N. Balb/c mice (groups of five animals) were subcutaneously immunized three times every 2 weeks with a 1:1 mixture of 100 µg of NE or PEs (Pp-N, Bv-N, Pp-P1 and Bv-P1) and aluminum hydroxide (Croda, Snaith, UK) adjuvant–PBS per dose. Immunization with an equivalent volume of adjuvant solution was used as a control. At 7 days after the final dose, the blood and spleens were collected for subsequent assays. Specific murine antibodies against NEs were detected using ELISA as described above, with minor modifications. Antigen coating and blocking were performed as described above. For total IgG, IgG1, and IgG2a, mouse sera were diluted 1:2,000. For specific IgE detection, mouse sera were diluted 1:4. Detection of specific antibodies was achieved using 1:2,000 HRP-labeled antibodies (anti-mouse IgG, IgG1, IgG2a, or IgE from Sigma-Aldrich) and OPD substrate as explained.

The IL-10 measurement followed the methodology previously described by our group (31). Spleens were isolated aseptically, minced, and filtered through sterile 100-µm steel filters. Erythrocyte-free cells suspended in PBS containing 0.5% fetal bovine serum (FBS; Gibco) were stained with 2 μM carboxyfluorescein succinimidyl ester (CFSE) for 10 min in the dark. Afterward, the cells were washed with PBS containing 2% FBS. CFSE-stained cells were cultured into 48-well plates at a concentration of 2 × 10⁶ cells/well in the presence of 5 or 50 µg/ml Pp-N or Bv-N extract. Cells cultured with 2.5 μg/ml of concanavalin A (Sigma-Aldrich) were used as a positive proliferation control. Cells without stimulant were used as the negative control. After 5 days, proliferation was measured on an FC-500 flow cytometer (Beckman Coulter, Brea, CA, USA). IL-10 was measured in 48-h culture supernatants using a specific ELISA quantification kit (R&D Systems). Data were analyzed using the FlowJo v.10 software and GraphPad Prism software.

Data were analyzed with GraphPad Prism 10 software. For the analysis of MAT, comparisons among conditions were performed using two-way ANOVA followed by Dunnet’s test. Data were considered significant at p < 0.05.

For the analysis of the antibody levels in immunized mice, lymphoproliferation, and IL-10 measurements, comparisons for non-parametric data were performed using the Kruskal–Wallis test, while that for parametric data was performed using one-way ANOVA.

For exploration and visualization of the correlation between all variables examined in the physicochemical, molecular, and biological behavior of polymers, we used principal component analysis (PCA) to statistically aggregate these variables, reducing the number of observed variables into a smaller number of principal components (PCs). For correct aggregation of the variables, the data from Pp-N-derived samples were normalized relative to the highest value in each group, and the same was done for the Bv-N-derived samples. PCs (PC1 and PC2) were selected with the largest eigenvalues that together explained 90% of the total variance. The results are displayed in a PC1 versus PC2 graph.

Five PEs were generated from each NE by decreasing (0.1× and 0.2×) or increasing (5× and 10×) the GA concentration relative to the reference condition (1× = 25 mM, P1). The resulting preparations are denoted Pp P10 → Pp P0.1 and Bv P10 → Bv P0.1 for P. pratense and B. verrucosa, respectively.

The polymerization efficiencies, which were calculated from the residual LMWFs (Table 1), ranged from 5% (P10) to 97% (P1). SDS-PAGE analysis confirmed successful polymerization, with all PEs displaying a predominant band above 250 kDa, while the NEs showed multiple bands from 10 to 250 kDa (Figures 1A, B, top). A smear appeared only in the LMWF of P10 and P5 (Figures 1A, B, bottom) and in the glycine–GA control reactions (Supplementary Figure 2), suggesting the presence of glycine–GA adducts.

One-dimensional ¹H NMR spectroscopy further supported the formation of HMWF PEs, as evidenced by line broadening and the signal attenuation in all PEs compared with their respective NEs (Figures 1C, D), consistent with extensive cross-linking. The polymer stability over time was assessed through longitudinal NMR measurements. By comparing the spectral signal intensities at multiple time points (T0 → T3), degradation was monitored as described (25). NEs showed progressive signal loss indicative of degradation, whereas PEs remained largely stable. Notably, Pp-P1 and Bv-P1 showed the highest stability over the 4-month period (Figures 1E, F).

The hydrodynamic size distribution of the polymers was examined using two orthogonal approaches: DOSY-NMR and TEM. The apparent MW distribution derived from DOSY (Supplementary Figure 3) ranged from 100 to 700 kDa for P. pratense PEs (Figure 1G) and from 90 to 300 kDa for B. verrucosa PEs (Figure 1H). In both cases, P1 extracts exhibited the largest average molecular sizes. These findings were corroborated by TEM, which showed that P1 samples formed the largest structures, while extreme low GA conditions (0.1×) led to smaller, fragmented morphologies (Figures 2A–D). In addition, we attempted to validate these results using DLS on Bv-N-derived PEs (Supplementary Figure 4). DLS resolved only four to six broad populations with high polydispersity indices, each likely encompassing multiple particle sizes, suggesting that this technique does not have sufficient resolution to accurately resolve the high degree of polydispersity present in our samples. DLS has limited resolution toward large species with hydrodynamic radii approaching or exceeding the micrometer range, which further restricts the reliable characterization of these samples by this technique.

Morphological and porosity differences were evident in the TEM analysis. While NEs appeared as irregular and dispersed particles, possibly representing protein aggregates, PEs prepared at moderate to high GA ratios (e.g., Pp-P1, Pp-P5, Bv-P1, and Bv-P5) formed dense or filamentous networks. In contrast, low and high GA concentrations (e.g., Pp-P0.1 and Pp-P10) produced smaller, less interconnected structures (Figures 2A, B). Image-based porosity analysis revealed the following increasing order of porosity (Figures 2E, F):

Together, these results suggest that GA stoichiometry modulates the key structural parameters of PEs, including the yield, size, morphology, porosity, and stability.

To determine which allergens remained detectable following GA polymerization, the NEs and PEs were analyzed using LC-MS/MS and immunoblotting.

Proteomic profiling by LC-MS/MS was performed on three representative polymerization conditions from each pollen source (P10, P1, and P0.1) (Figures 3A, B). For P. pratense, all examined PEs (Pp-P10, Pp-P1, and Pp-P0.1) contained the major allergens, apart from Phl p 12 and Phl p 13. In contrast, allergen recovery from B. verrucosa was more variable: only Bv-P1 retained the full panel of known allergens, except for Bet v3, which was absent from all PEs. In addition, several proteins were detected uniquely in the PEs, but not in their corresponding NEs (Figures 3C, D), suggesting altered detectability of the high-molecular-weight species following cross-linking. A summary of these proteins is provided in Supplementary Table 1 for the Pp-derived samples and in Supplementary Table 2 for the Bv-derived samples, while the corresponding raw datasets are available in Supplementary data 1–4 (for Pp) and Supplementary data 5–8 (for Bv).

WB analysis corroborated the MS findings and provided additional insights into IgE and IgG epitope availability. In P. pratense (Figure 3E), residual IgE binding at >250 kDa was observed across all PEs. However, polyclonal IgG from mice immunized with Pp-N (Figure 3E) or Phl p 5 (Supplementary Figure 5A) only lightly cross-linked the Pp-P0.2 and Pp-P0.1 samples, implying that stronger cross-linking (≥P1) leads to epitope masking or modification. Importantly, no IgG reactivity was detected in the LMWFs, confirming effective polymerization (Figure 3E; Supplementary Figure 5A).

For B. verrucosa, IgE and murine IgG reactivity specific for Bv-N and Bet v 1 were consistently detected at >250 kDa in all polymerized samples. Notably, Bv-P10 exhibited low IgE and IgG signal intensities, whereas Bv-P0.1 showed strong IgE but weak IgG binding (Figure 3F; Supplementary Figure 5B). The intermediate conditions (Bv-P5, Bv-P1, and Bv-P0.2) displayed similar binding profiles for both antibody types.

In some LMWFs, discrete immunoreactive bands persisted: a 28-kDa IgE binding band in Pp-P0.1 and a 17-kDa band in Bv-P0.2 and Bv-P0.1 (Figures 3E, F; Supplementary Figure 5B), likely representing unpolymerized monomeric allergens that escaped cross-linking.

Taken together, these results show that GA polymerization retains the majority of the clinically relevant allergens.

To evaluate the effect of GA cross-linking stoichiometry on allergenicity, we assessed the IgE binding and effector cell activation using a series of established immunoassays.

All PEs displayed reduced IgE binding capacity compared with their respective NEs, as determined with indirect ELISA (Figures 4A, B). This reduction was statistically significant for all preparations, except those produced with the lowest GA concentrations (i.e., Pp-P0.1, Bv-P0.1, and Bv-P0.2).

Consistent with these findings, the ELISA inhibition assays showed a stoichiometry-dependent reduction in BP. At a concentration of 2 µg/ml, Pp-N achieved 97.8% inhibition of IgE binding (BP = 0.198), while the most extensively cross-linked Pp-P10 only inhibited 39.8% (BP = 364.9) (Figure 4C and Table 1). A similar trend was observed for B. verrucosa, with Bv-N showing complete inhibition (BP = 0.067) and Bv-P0.1 exhibiting only 41% inhibition (BP = 112.1) (Figure 4D and Table 1), indicating a diminished capacity of the PEs to block IgE recognition of NEs.

Although the overall KSCN elution profiles were comparable across extracts (Supplementary Figure 6), the avidity indices, used here to estimate the strength of IgE–epitope interactions, showed measurable differences (Table 1). For P. pratense, the index declined from 1.087 in Pp-N to 0.800 in Pp-P0.2; for B. verrucosa, it dropped from 0.960 in Bv-N to 0.725 in Bv-P0.1. These changes reflect a weakening of the IgE binding strength associated with specific GA-to-protein ratios.

The functional consequences of these differences were evaluated by MAT. The lowest activation levels (expression of the CD63 and CD107a surface markers and β-hexosaminidase release) were observed for the most extensively cross-linked P10, P5, and P1 formulations, both for P. pratense-derived (Figures 4E–G; Supplementary Figures 7A, C–E) and B. verrucosa-derived PEs (Figures 4H–J; Supplementary Figures 7B, F–H). In contrast, P0.2 and P0.1 induced mast cell responses comparable to those elicited by the NEs.

These results suggest that GA-mediated polymerization reduces both IgE binding and effector cell activation in a stoichiometry-dependent manner.

To better understand how structural and immunological properties are interrelated across the different polymerization conditions, PCA was performed using all measured physicochemical, molecular, and biological parameters (see Section 3.2). The resulting biplot (Figure 5) revealed three distinct clusters that corresponded to specific GA cross-linking stoichiometries.

The first cluster was defined by high mast cell activation and elevated IgE binding and included samples generated at the lowest GA-to-protein ratios (Pp-P0.1, Bv-P0.1, and Pp-P0.2). This group reflected the least effective cross-linking, associated with poor allergen masking and high effector cell activation.

The second cluster comprised samples with high IgE binding, avidity, and low polymerization yield. It was associated with the highest GA concentrations (Pp-P10 and Bv-P10). These results suggest that, while excessive GA results in dense structures with strong IgE–epitope interactions, it may also limit the efficiency of polymer formation.

The third cluster included preparations with large sizes, high density, and reduced allergenicity and was centered around the reference GA-to-protein ratios (Pp-P1 and Bv-P1). These samples consistently exhibited the most favorable structural and immunological profiles, supporting their selection as optimal formulations for further immunological evaluation.

Interestingly, BP, as assessed using ELISA inhibition, did not cluster with any of the other variables. This indicates that BP may be influenced by additional factors not fully captured by the parameters included in this analysis.

To evaluate whether the most hypoallergenic polymers (P1) retained immunogenicity, BALB/c mice were immunized with P1 PEs or their corresponding NEs.

Both PEs and NEs from P. pratense and B. verrucosa elicited comparable levels of allergen-specific IgG (Figures 6A, B), including the IgG1 and IgG2a isotypes (Supplementary Figures 8A, B, E, F), indicating preserved humoral responses. All groups showed detectable IgE titers, with no statistically significant differences between the PEs and NEs (Supplementary Figures 8C, G). Notably, Bv-P1 induced specific IgE levels that were indistinguishable from the adjuvant-only control. Evaluation of the IgG2a/IgE ratio revealed similar values for Bv-N and Bv-P1, while Pp-P1 showed a slightly lower, but not statistically significant, ratio compared with Pp-N (Supplementary Figures 8D, H).

Cellular immune responses were also maintained. Splenocytes from P1-immunized mice showed proliferation (Figures 6C, D) and IL-10 production (Figures 6E, F) upon stimulation with NEs at levels comparable to those observed in NE-immunized animals, but significantly higher than the negative control, supporting the preservation of T-cell reactivity.

Taken together, these results demonstrate that P1 polymers, despite their reduced allergenicity and high structural density, retain the immunogenic profile of NEs.