Non-specific lipid transfer proteins were purified from apple (Malus domestica, Mal d 3), peach (Prunus persica, Pru p 3) and hazelnut (Corylus avellana, Cor a 8) as described previously ^[20]–^[22]. 7S and 11S seed storage globulins were purified from hazelnut (Corylus avellana, Cor a 11, Cor a 9) and peanut (Arachis hypogaea, Ara h 1, Ara h 3) as described previously ^[22], ^[23]. Caseins were purified from cows' (Bos domesticus, Bos d 8) and goats' milk (Whole caprine casein, WCC ^[24]). Tropomyosin, the major allergen from brown shrimp (Penaeus aztecus, Pen a 1) was produced as a recombinant protein and purified accordingly ^[17]. The natural and recombinant allergens were tested for their IgE-binding activity using allergic patients sera as previously described and their primary and secondary structures verified by N-terminal sequencing, MALDI-TOF-MS analysis and far-UV CD-spectroscopy ^[5]. An overview of the allergens used in this study, their acronyms and physicochemical parameters together with the experimental conditions of the NMR analyses are summarised in Table 1.

Allergen solutions were prepared in H₂O/D₂O 9∶1, (0.50 ml, with concentration ranging from 0.39 mM to 0.03 mM) in aqueous buffer systems (pH ranging from 7.0 to 7.8) as detailed in the legends for figures. The solutions were transferred into high-quality NMR tube with argon as headspace gas. Before and after the NMR experiments, the allergens were stored at 253 K (−20°C). Two High Resolution NMR experiments were carried out, using an Avance 700 spectrometer by Bruker, operating at a proton resonance frequency of 700 MHz (11.7 Tesla). The two experiments differed in the method of managing the water signal. The pulse program of the first experiment included a 1D sequence with pre-saturation to minimize the water signal ^[25]. The second experiment is based on gradient-based water suppression by 1D excitation sculpting using 180° water-selective pulses ^[25]. It increases the overall sensitivity and decreases the distortion of the baseline, but some signal near the (suppressed) water peak can be suppressed. This can be checked by the pre-saturation experiment.

For each sample and for each experiment 256 scans at 298 K (25°C) were routinely programmed, due to the low sample concentration. More scans were required for proteins with higher molecular mass and more diluted protein solutions (Mal d 3, Ara h 3/4, Cor a 9, Ara h 1, Cor a 11) as specified in the legends of the corresponding figures, where variations in the number of scans and of the field strength are reported, with the molecular weights and the molarities of the samples and the buffer conditions. The spectra were initially evaluated considering the dispersion of the signals in the regions of the methyl protons (−0.5–1.5 ppm), alpha-protons (3.5–6.0 ppm), and amide protons (6.0–10.0 ppm).

The consistency of the allergen spectra with information available in the scientific literature and with structures either deposited into the Protein Data Bank (PDB) or supported by a publication, was routinely checked. If no experimental structure was available, the comparison with a modelled structure was performed ^[26]–^[29] and its consistency with the NMR spectra checked.

The stability of the nsLTPs (Mal d 3, Pru p 3 and Cor a 8) to thermal stress was studied using a 600 MHz spectrometer equipped with a probe suited to work at high temperature. A series of experiments was performed consisting of stepwise heating of each sample from 298 K (25°C) to a maximum 358 K (85°C), by 5 K intervals. After each step an NMR water pre-saturation (zgpr) experiment was launched and the spectrum evaluated. If significant structural changes were detectable, the sample was cooled to 298 K and scanned to verify whether the detected structural change was reversible. If the maximum temperature allowed by the probe had not been already reached, the sample was heated to the next threshold.

A minimum of 128 scans was programmed for each NMR experiment, extended to 512 for the most diluted samples (Mal d 3). A minimum of 20 minutes was required to reach the thermal equilibrium after each 5 K heating step. A minimum of 90 minutes was required after each cooling step and after each following heating step. All the experiments were conducted under near-neutral pH condition. The thermal stability of Pru p 3 was also studied in acidic solution (adjusting to pH 3 by addition of HCl).The pH was checked before and after the experiments.

The modelled structures were obtained using the fully automated protein structure homology-modelling server SWISS-MODEL ^[25], ^[27], ^[28] and were selected based on the protein family of the template, the sequence identity between the target and the template, the modelled residue range of the target, the QMEAN Z-Score and the QMEAN4 global score of the model ^[26], ^[30]–^[32]. Models of low quality are expected to have strongly negative QMEAN Z-scores. The QMEAN describes the likelihood that a given model is of comparable quality to experimental structures. The QMEAN4 ranges between 0 and 1 with higher values indicating better models among alternative models of the same target.

Since large chemical shift dispersions, especially in the methyl group region between 1.0 ppm and −1.0 ppm and in the amide region down-field of 8.5 ppm can be used as a reliable indicator for protein folding ^[10], ^[33] this approach was taken to assessing the structural attributes of the selected allergens. Few broad peaks around 8.3 ppm are a good indicator of a protein without an ordered tertiary structure, because this is a region characteristic of backbone amides in random-coil configuration. By contrast, several peaks, possibly narrow and resolved, dispersed beyond 8.5 ppm (8.5–11 ppm) are the signature of a rigid and ordered tertiary structure. However, the signals of the amide protons can be broadened due to chemical exchange especially at non-acidic pH. This can be a drawback at protein concentrations close to the detection limit when amide signals are to be observed. Isolated methyl group signals upfield of 0.5 ppm can be more sensitive and reliable probes for folded protein as they are not affected by chemical exchange broadening. In addition, their signal intensities are higher because three equivalent protons contribute rather than one proton of the amide group. Therefore, a wide signal dispersion in the aliphatic region of the spectrum between 1.0 and −1.0 ppm, versus a steep flank of the dominant peak at approximately 1 ppm reliably discriminates rigid and ordered tertiary structures from random-coil or flexible structures. This is because the loss of high field methyl peaks indicates a reduction in tertiary side chain organization, which can be considered complete when the NMR spectrum is characterized by few, rather narrow, but poorly dispersed peaks that are indicative of extremely mobile residues. This condition can be usually achieved as a result of protein unfolding in strongly acidic condition usually in the presence of a high concentration of urea ^[34].

A thermodynamically distinct state of proteins showing properties of both native and denatured proteins is “molten globule” ^[35]. Molten globules are considered as intermediates on the protein folding/unfolding process but in some cases, molten globules exist as stable conformers, with rigid but different tertiary structure ^[9]. The NMR spectrum of a molten globule protein is characterized by few, extremely broad peaks, especially for most of the backbone amide groups.

Whilst spectral signals from certain allergens (Ara h 1, Cor a 9, Cor a 11, Mal d 3) were weak and spectra were dominated by buffer signals this did not affect the readability of the critical spectral regions.

All the spectra were consistent with the corresponding structures deposited in the PDB, or with structures that could be modelled starting from the sequence and anticipated on the basis of the proteins family membership. Analysis of the 1D ¹H-NMR spectra then allowed the classification of the allergens under study as follows.

The NMR spectra of these nsLTPs (Fig. 1 and Fig. 2) show the unmistakable features of rigid and extended tertiary structures which is not affected by low pH environment and is in line with the 2D studies performed with Pru p 3 by Wijeshina-Bettoni et al ^[14]. There are many narrow and resolved peaks throughout the whole spectral window, notably in the methyl and amide-aromatic regions. The spectra of these allergens are very similar: for example, the extreme methyl region between 1 and 0 ppm and the aromatic region between 6 and 8 ppm. This fact gives experimental support to the similarity of the modelled structures of Mal d 3 and Cor a 8, and the experimental structure of Pru p 3 (Fig. 3), as expected from the high degree of sequence homology and conserved cysteine residues forming disulfide bridges. Thus Mal d 3 (Uni Prot entry: Q9M5X7), and Cor a 8 (Uni Prot entry: Q9ATH2), share 81.3% and 59% sequence identity with Pru p 3 (Uni Prot entry: P81402), respectively. The overall intact 3D structure is relevant for IgE-binding activity of Pru p 3 and is reduced when alkylation and reduction of the protein was performed ^[36]. Therefore, Pru p 3 was subjected to stepwise heat treatment to investigate whether heating processes have an impact on its overall structure.

During thermal treatment the spectra gradually change towards the denatured state. Under acidic conditions (Fig. 4, pH 3) the increasing loss of tertiary structure of Pru p 3 is apparent from the decreasing dispersion of peaks either in the methyl region (around 0 ppm) or in the amide-aromatic region (6–10 ppm). Irreversible denaturation is not detectable until 358 K (85°C) and the 298 K (25°C) spectra, scanned before and after heating, overlap perfectly. These data are in good agreement with the studies on heat resistance of barley LPT from Perrocheau et al. ^[37]. Under neutral conditions (Fig. 4, pH 7) two major differences can be observed during thermal denaturation of natural Pru p 3. The former is an irreversible structural transition occurring between 353 K (80°C) and 358 K (85°C): the final room temperature spectrum shows only a few broad and poorly dispersed peaks in the amide-aromatic and in the methyl region and the absence of resonances upfield and downfield. The latter represents the disappearance of amide resonances (7–9 ppm) as the temperature rises. This is due to proton exchange with the solvent, and depends on pH, temperature and accessibility of exchangeable protons, which increases with unfolding, and can be due to the transition to a molten-globule state. Native Mal d 3 shares the same behaviour as Pru p 3 when heated in neutral conditions (data not shown). Heating was stopped at 348 K (75°C) for Cor a 8 due to instrumental problems. The protein refolds to the 25°C tertiary structure after the heat-induced unfolding at the maximum tested temperature (data not shown).

The higher molecular mass of globulins as compared to nsLTPs and their propensity to aggregate into more complex quaternary structures seriously limits the resolution of their NMR spectra (Fig. 5). This could be a particular problem for Cor a 9, due to the possible presence of the hexameric form. A series of signals, though basically unresolved, can be observed in the extreme methyl region of all the spectra (around 0.5 ppm) indicating regions of rigid tertiary structure. On the other hand, the presence of extended mobile regions must be considered, due to the few broad peaks that can be observed in the whole spectral window, notably in the HN region between 6 and 8 ppm. By contrast, the same region of the Cor a 11 spectrum is populated by several better resolved and narrow resonances that give a clear indication of rigid and ordered regions. Both, Cor a 9 and Cor a 11 show a noticeable signal broadening in the backbone NH region between 8.0 and 8.7 ppm, that could indicate the presence of a molten globule state. Narrow, weak signals are observed against a background of broad peaks in the Ara h 3/4 spectrum which could be due to the presence of both lighter monomers and more massive aggregations in the mixture (Fig. 5).

Up to now two sequences of 7/8S globulin Ara h 1, from peanut (63.5 kDa) (UniProt entries P43237 and P43238) were characterised ^[38]. Very recently, the structure from the Ara h 1 core region without the N-terminal extension and the C-terminal flexible region was determined ^[39]. Both isoforms could be modelled using the 7S globulin from adzuki bean ^[40] as a template (PDB accession 2ea7), with 48% of sequence identity (Fig. 6). The template and the modelled monomer can be divided into two similar modules (N- and C-terminal), each showing an extended loop domain. The models could be verified by the crystal based analysis of the Ara h 1 core fragment ^[39], ^[41]. Already identified IgE epitopes, were mapped on the models and it became evident that some of these epitopes were only surface exposed when the trimer formation was disrupted ^[42].

A partially resolved structure (PDB entry 3c3v) has been reported as Ara h 3 (now designated: Ara h 3.0101; Swiss prot Entry: O82580) ^[43]. It consists of an N-terminal domain and a C-terminal domain. Three regions of the N-terminal domain and the tail of C-terminal domain are disordered (Fig. 7). Using this structure as a template a model was built of the isoform reported as Ara h 4 (now designated: Ara h 3.0201; Swiss Prot entry: Q9SQH7) ^[44]. The model which shares 75.5% sequence identity with “Ara h 3” (Fig. 7) was calculated including the unresolved regions of the experimental structure, and is consistent with the presence of extended, partially structured, mobile regions.

IgE epitopes were identified on the surface of Ara h 3, some of them specific for the peanut allergen and some of them shared with other legumin type allergens highlighting the molecular basis for cross-reactivity ^[45].

No experimental structure of hazelnut globulins is known to date. Again a homology model was built (Fig. 6) using the Cor a 11 sequence ^[46] starting from its 48 kDa precursor sequence (Swiss Prot entry: Q8S4P9), with 32.5% of identity with the template PDB 2ea7, chain A, corresponding to the adzuki bean 7S globulin ^[40]. The model is trimeric, with each monomer formed by two very similar modules and each module containing an extended loop domain. Similarly a model was built using the Cor a 9 sequence (GenBank: AAL73404.1) which has 51.8% sequence identity to the almond globulin ^[47], ^[48] (PDB 3fz3) (Fig. 8). The structure of this template, and of the derived model, shows similarities with the structure of the 11S globulin Ara h 3/4 (Ara h 3.0101 and Ara h 3.0201; Fig. 7 and Fig. 8).

Narrow, weak signals in are observed against a background of broad peaks in the Ara 3/4 spectrum which could be due to the presence of both lighter monomers and more massive aggregations in the mixture (Fig. 5).

The spectra of all of the globulins provide experimental support for their modelled structures and are consistent with our knowledge of the structural characteristics of this family of proteins including their tendency towards aggregation, the ubiquity of isoforms, and the presence of extended mobile regions in their tertiary structure.

The NMR spectra of the sample of Pen a 1 under study and the spectra of caseins are very similar. These spectra (Fig. 1) show the typical characteristics of random coil structures with extended regions of the spectral window being devoid of signals and a few broad peaks at characteristic intervals. The dominant methyl peak at approximately 1 ppm is a distinctive signature of a random-coil or extremely flexible structure. Single narrow signals can be observed on the envelope of broad peaks which are not resolved to the baseline because they are confined in narrow spectral ranges, confirming the high mobility of the residues of these samples. A model was constructed of Pen a 1 (Accession No. DQ151457.1) ^[49] based on the structure of pig cardiac muscle tropomyosin ^[50] (PDB 1c1g, chain A) with which it shares 56.2% sequence identity (Fig. 9). This has a coiled-coil structure which can undergo a transition from a coiled coil to molten-like state before complete dissociation of the two chains, chains that are flexible and can be unfolded at physiological temperature ^[51]. This could explain the observation that the NMR spectrum is characteristic of a mobile, unfolded structure. Detailed studies on the epitope mapping of this major allergen identified linear IgE epitopes with increased antibody binding activity when grafted onto non allergenic tropomyosin ^[17].

The spectra of the caseins were similarly characterised by a lack of signals between 6.2 and 6.6 ppm and 0.0 and 0.7 ppm, respectively (Fig. 1) indicative of their highly mobile structure. The fact that unfractionated caseins are mixtures of four different molecules is not apparent because alpha-s1-casein, beta-casein, alpha-s2-casein and k-casein ^[19], all share similar molecular weight and lack secondary and tertiary structure, and therefore their residues share very similar structural environment as confirmed also by the great similarity of all these spectra.

On the basis of the NMR analysis the proteins could be classified into three groups. The first group included those molecules whose spectra showed the unquestionable signature of their tertiary structure: nsLTPs (Fig. 1 and Fig. 2) and the hazelnut 7/8S globulin, Cor a 11. However, the NMR data indicate that the latter may contain disordered stretches (Fig. 5) as has been observed for seed storage globulins from other plant species such as soybean ^[52].

The second group included those allergens whose spectra clearly showed no fingerprints for tertiary structure (Fig. 1). These were whole caseins from cows' and goats' milk (Bos d 8, WCC) and tropomyosin from brown shrimp (Pen a 1). The spectra for caseins are consistent with other observations of protein mobility which has led to them being termed rheomorphic ^[19].

The third group included allergens whose spectra revealed an intermediate structure: the 11S globulin Cor a 9 and the 7/8S globulin Ara h 1. Their spectra (Fig. 5) indicated that the protein could be structured with well organized rigid regions with an ordered tertiary structure, interspersed with disordered regions (as indicated by temperature factors in the x-ray structures of certain seed storage globulins and the fact certain regions had not been resolved and appear to be highly mobile even in a crystal lattice ^[41]. Alternatively these data indicate the sample could be a mixture of folded and unfolded or differently folded proteins. The spectrum of the 11 S globulin Ara h 3/4 (Fig. 5) also indicated the presence of both disordered and ordered parts.

It is worth noting that some of the allergens investigated displayed molecular masses, (e.g. 60 kDa without considering possible aggregations) that can be considered beyond the limitations of the method and the instruments adopted.

High quality allergen batches with defined physicochemical features are essential for reliable and sensitive in vitro diagnosis for allergy. The 3-dimensional structures of the most important allergen families have important impacts on their IgE binding activity. It has become evident that food processing can affect the structure and thus allergenicity. Therefore, methods to assess the structural integrity of purified allergens are urgently needed. The present study has shown that 1D NMR can be used to generate a fingerprint containing the overall structural and dynamic information of allergenic proteins in a high throughput manner.