High-throughput NMR assessment of the tertiary structure of food allergens

Abstract

Background: In vitro component-resolved diagnosis of food allergy requires purified allergens that have to meet high standards of quality. These include the authentication of their conformation, which is relevant for the recognition by specific IgE antibodies from allergic patients. Therefore, highly sensitive and reliable screening methods for the analysis of proteins/allergens are required to assess their structural integrity. In the present study one-dimensional 1H Nuclear Magnetic Resonance (1D 1H-NMR) analysis was adopted for the assessment of overall structural and dynamic properties and authentication of a set of relevant food allergens, including non-specific lipid transfer proteins from apple, peach and hazelnut, 7/8S seed storage globulins from hazelnut and peanut, 11S seed storage globulins from hazelnut and peanut, caseins from cows' and goats' milk and tropomyosin from shrimp.

Methodology/principal findings: Two sets of 1D 1H-NMR experiments, using 700 MHz and 600 MHz instruments at 298 K were carried out to determine the presence and the extent of tertiary structure. Structural similarity among members of the individual allergen families was also assessed and changes under thermal stress investigated. The nuclear magnetic resonance (NMR) results were compared with structural information available either from the literature, Protein Data Bank entries, or derived from molecular models.

Conclusions/significance: 1D (1)H-NMR analysis of food allergens allowed their classification into molecules with rigid, extended and ordered tertiary structures, molecules without a rigid tertiary structure and molecules which displayed both features. Differences in thermal stability were also detected. In summary, 1D (1)H-NMR gives insights into molecular fold of proteins and offers an independent method for assessing structural properties of proteins.

Figure 1. NMR spectra typical for allergens with random coil structure, compared with a spectrum of an allergen with rigid tertiary structure.

Two enlarged regions of the spectra (gradient-based water suppression NMR experiment) of purified casein fraction from cow (Bos d 8, 0.08 mM, 900 MHz cryoprobe, 40 mM phosphate, pH 7.4), and goat (whole caprine casein, WCC, 0.19 mM; 40 mM phosphate pH 7.4), and of the recombinant tropomyosin from shrimp (Pen a 1, 0.13 mM, 36.4 kDa, MOPS 20 mM, NaCl 0.5 M, pH 7.6) are shown. The 0.3–1.2 ppm region includes the extreme methyl signals while the 6.2–7.4 ppm region includes the resonances of aromatic ring protons, side chain HN, and part of backbone HN; These spectra are typical of random coil structures, as indicated by the absence of signal upfield 6.6 ppm and 0.7 ppm. The same enlarged regions of the spectrum of Pru p 3 (0.39 mM, 91 AA, 9.2 kDa) are shown (top) to underline the difference between spectra of folded and unfolded proteins.

Figure 2. NMR Spectra of non specific lipid transfer protein, nsLTP, from peach (Pru p 3), apple (Mal d 3) and hazelnut (Cor a 8).

The spectra (gradient-based water suppression NMR experiment) of nsLTPs from different sources are compared: Cor a 8 from hazelnut (115 AA, 11.8 kDa, 0.15 mM; 25 mM phosphate, NaCl 150 mM, pH 7.0), Mal d 3 from apple (115 AA, 11.4 kDa, 0.08 mM, 6504 scans; 20 mM phosphate, pH 7.4), Pru p 3 from peach (91 AA, 9.2 kDa, 0.39 mM; 25 mM phosphate, pH 7.0). These spectra indicate the presence of extended, rigid and ordered tertiary structure for all three molecules and display a high degree of similarity.

Figure 3. Modelled and experimental structures of nsLTPs from apple, peach and hazelnut.

Models of Mal d 3 and Cor a 8 based on the experimental structure of Pru p 3 as template (PDB 2alg, chain B, 92 residues). Disulphide bonds are rendered as sticks and the internal hydrophobic cavity is rendered as a mesh surface. The modelled residue range is 25–115 for both models. QMEAN Z-score = −1.895 (Mal d 3) and −1.639 (Cor a 8).

Figure 4. NMR spectra of nsLTP from peach, Pru p 3 under thermal treatment in acidic and near-neutral pH conditions.

Heating of Pru p 3 (0.39 mM, 91 AA, 9.2 kDa) under acidic (pH 3.0) and neutral (pH 7.0) conditions; pre-saturation NMR experiments. To improve readability, the spectra at T>298 K are aligned with the reference spectrum at 298 K (25°C), by shifting them upfield as indicated after the temperature. pH 3.0 bottom: NMR spectra of Pru p 3 scanned at 298 K before and after heating up to 358 K (85°C). pH 7.0, bottom: NMR spectra of Pru p 3 scanned at 298 K before and after heating up to 358 K. top: spectrum scanned at 298 K after 358 K heating. In neutral conditions the protein undergoes irreversible denaturation, in the 80°C–85°C range of temperature.

Figure 5. NMR Spectra of globulins from peanut and hazelnut.

The spectra (gradient-based water suppression NMR experiment, 848 scans) of selected globulins from different sources are presented: Ara h 3/4 (mixture of isoforms of 11S globulins from peanut, 37.0 kDa, putative concentration 0.14 mM; 50 mM phosphate, NaCl 150 mM, pH 7.5), Cor a 9 (11S globulin from hazelnut, 40 kDa, 0.03 mM, 50 mM phosphate, NaCl 150 mM, pH 7.5), Ara h 1 (7/8 S globulin from peanut, 63.5 kDa, 0.03 mM, 50 mM MOPS, NaCl 200 mM, pH 7.8), Cor a 11 (7/8S globulin from hazelnut 48 kDa, 0.04 mM, phosphate 50 mM, NaCl 150 mM, pH 7.5), Signals, though basically unresolved due to high MW and possible aggregation, can be observed in the extreme methyl region of all the spectra (around 0.5 ppm) which indicate of regions of rigid tertiary structure. The presence of extended disordered regions must also be considered, due to the few broad peaks that can be observed in the whole spectral window, especially in the HN region between 6 and 8 ppm and especially for Cor a 9. On the contrary the same region of Cor a 11 is populated by several more resolved and narrow resonances, indicative of rigid and ordered regions. Both, Cor a 9 and Cor a 11 show a noticeable signal broadening in the backbone NH region (8.0–8.7 ppm), indicating the presence of the molten globule state in solution. In the Ara 3/4 spectrum, narrow, weak signals, can be observed on an envelope of broader peaks. They could be due to the lighter (non-aggregated) component of the mixture.

Figure 6. Modelled and experimental structures of 7/8S globulins from peanut, hazelnut and adzuki bean.

Models of 7/8S globulins from peanut (Ara h 1, 614 residues) and hazelnut (Cor a 11, 448 residues), both based on 7S globulin from adzuki bean (PDB 2ea7 chain A) as template. Either the monomer (chain A) or the trimer (chain A rendered with the same orientation, in light gray) of the template are shown. The presence of mobile parts is indicated by the many extended loops of the models, corresponding to similar extended loops or to unresolved regions of the template. Modelled residue range: 161–583 (Ara h 1), 55–435 (Cor a 11). Sequence Identity [%]: 49 (Ara h 1), 32 (Cor a 11). QMEAN Z-score = −1.614 (Ara h 1), −0.956 (Cor a 11). QMEAN Z-score of Ara h 1 = 0.669 (0.600 if it is modelled on PDB 2cv6 chain A, corresponding to 8S globulins of mung bean).

Figure 7. Modelled and experimental structures of Ara h 3/4 isoforms.

Experimental structure of Ara h 3/4 (PDB 3c3b) and model of its sequenced isoform (Q9SQH7 GenBank AAD47382.1; 530 residues) based on PDB 3c3v as template. The presence of mobile parts is indicated by the many extended loops of the model, corresponding to similar extended loops or to unresolved regions of the template. Disulphide bonds are rendered as sticks. Modelled residue range: 21–521. Sequence Identity [%]: 75. QMEAN Z-score = −1.933.

Figure 8. Modelled and experimental structures of 11S globulins from hazelnut and almond.

Model of Cor a 9 (11S globulin from hazelnut), based on template PDB 3fz3 (11S globulin from almond) chain C (515 residues). Disulphide bonds are rendered as spheres. Both the monomer and the hexamer of the template are shown. The template's chain C is rendered in light gray, with the same orientation, either as monomer or in the hexamer. The presence of mobile parts is indicated by the many extended loops of the model, corresponding to similar extended loops or to unresolved regions of the template. Modelled residue range: 34 to 496. Sequence Identity [%]: 51. QMEAN Z-score = −3.084.

Figure 9. Structures of tropomyosin from brown shrimp (modelled) and pig (experimental).

Model of Pen a 1, the tropomyosin from brown shrimp (284 residues) based on tropomyosin (PDB 1c1g chain A) from cardiac muscle of pig (Sus scrofa). Tropomyosins can pass from a coiled coil to molten-like structure before complete dissociation of the two chains. Chains are flexible and can be unfolded at physiological temperature. Modelled residue range: 1 to 283. Sequence Identity [%]: 56. QMEAN Z-score = −0.371. QMEAN score 4 = 0.755 (0.732 on tropomyosin from striated-muscle of R. norvegicus 2b9c chain A).