Solution structure of the strawberry allergen Fra a 1.

The PR10 family protein Fra a 1E from strawberry (Fragaria x ananassa) is down-regulated in white strawberry mutants, and transient RNAi (RNA interference)-mediated silencing experiments confirmed that Fra a 1 is involved in fruit pigment synthesis. In the present study, we determined the solution structure of Fra a 1E. The protein fold is identical with that of other members of the PR10 protein family and consists of a seven-stranded antiparallel β-sheet, two short V-shaped α-helices and a long C-terminal α-helix that encompass a hydrophobic pocket. Whereas Fra a 1E contains the glycine-rich loop that is highly conserved throughout the protein family, the volume of the hydrophobic pocket and the size of its entrance are much larger than expected. The three-dimensional structure may shed some light on its physiological function and may help to further understand the role of PR10 proteins in plants.

INTRODUCTION

In North and Central Europe as well as in North America approximately 15–20% of the population are allergic to pollen, with birch pollen being the main cause of those allergies [1,2]. Over 90% of the birch pollen allergic patients show sensitization to the major birch pollen allergen Bet v 1 [3]. In total 50–90% of the patients who are sensitized to Bet v 1 show IgE-mediated allergic cross reactions to food allergens such as those contained in fruits, vegetables and nuts [4,5]. These cross reactions are often based on the similarity of the three-dimensional allergen structures [6-9].

A series of Bet v 1 homologous food allergens have been identified and summarized in the superfamily of Bet v 1-like proteins, such as Api g 1 from celery [10], Mal d 1 from apple [11], Pru av 1 from cherry [12] and Gly m 4 from soya bean [13]. The members of the Bet v 1 superfamily are 153–160 amino acid proteins with molecular mass ~17 kDa and theoretical isoelectric points between pH 4.4 and 6.1. The sequence identities with Bet v 1 are 40–70%, and all these proteins show similar conformations with identical arrangements of secondary structure elements that correlate with the cross-reactivity.

The pollen and food allergens belong to the class 10 of pathogenesis-related proteins, protein family PR10, a subdivision of the Bet v 1-like superfamily. The PR10 proteins are conserved and ubiquitous in the plant kingdom, with roles in plant development [17] and immune defence induced by microbial attack [18], fungal elicitors [18,19], wounding [20] or stress stimuli [21]. The PR10 family contains, for example, the yellow lupine proteins LIPR-10.2A and LIPR-10.2B [14,15] and the St. John's Wort protein Hyp1 [16]. The classification of this protein group is based on the 3D (three-dimensional) structure. All structurally studied proteins show the same overall fold. However, their biological function is still not well recognized.

As an additional member of the Bet v 1 superfamily, a strawberry (Fragaria x ananassa) protein was detected in red fruited genotypes and named Fra a 1 [22]. A hypoallergenic white fruited genotype was almost devoid of the protein. Subsequently, five potential Fra a 1 isoforms (Fra a 1A–Fra a 1E) were identified with only seven variable amino acid positions, and it could be shown that Fra a 1A is recognized by IgE antibodies from strawberry allergic patients [23,24]. Two additional Fra a isoforms with greater variability were identified and named Fra a 2 and Fra a 3 [25]. Whereas Fra a 1e transcripts are expressed predominantly in roots with decreasing expression levels during fruit ripening, Fra a 2 expression increases in ripe fruits and Fra a 3 expression is constant in all stages. Transient RNAi (RNA interference)-mediated silencing of the Fra a genes in strawberries leads to reduced levels of anthocyanins and upstream metabolites as well as down-regulation of chalcone synthase (FaCHS) and PAL (phenylalanine ammonia lyase) genes [25], clearly indicating a key role of Fra a isoforms in the pigment synthesis pathway.

These different appearances of the members of the Bet v 1 superfamily of proteins, combined with the observation of divergent expression patterns in different species and different isoforms within a single species, renders it difficult to assign clear physiological functions to these proteins. One of the possible approaches to clarify the role of the Bet v 1 proteins in Nature is the study and comparison of their conformations in detail, an approach we here continue with the presentation of the solution structure of Fra a 1 in atomic resolution.

MATERIALS AND METHODS

Expression of recombinant Fra a 1e

The Fra a 1e gene was cloned into the expression vector pQE70 (Qiagen) by using the restriction sites SphI and BamHI. The protein contained a His6 tag at the C-terminus. The recombinant plasmid was transformed together with the vector pREP4 (Qiagen) into Escherichia coli BL21 (DE3). Fra a 1e was expressed in the host strain in LB (Luria–Bertani) medium at 20°C over 4 h. To obtain 15N- and 15N/13C-labelled protein overexpressing E. coli were grown on M9 minimal medium with 15N enriched (15NH4)2SO4 or 15N enriched (15NH4)2SO4/[U-13C]glucose respectively.

Protein purification

Crude cell extracts were prepared in 50 mM Hepes buffer (pH 7.0) with 300 mM NaCl for purification. Soluble His-tagged Fra a 1E was purified by immobilized metal-ion affinity chromatography [Ni-NTA (Ni2+-nitrilotriacetate) column with a bed volume of 1 ml; Qiagen] using a step gradient from 20 to 300 mM imidazole. Fractions were analysed by SDS/19% PAGE, and pure protein fractions were pooled and dialysed against 20 mM Hepes buffer, pH 7.0, containing 150 mM NaCl. This buffer was used as equilibration and elution buffer of a size-exclusion chromatography to remove residual impurities (Superdex S75 GL 10/300 column, total bed volume: 24 ml; GE Healthcare). Two litres of cell culture from LB and M9 minimal medium yielded approximately 11.5 and 7.6 mg of protein respectively. The purity of recombinant Fra a 1E exceeded 95% as estimated from SDS/PAGE on 19% gels. The DC protein assay (BioRad) as well as a molar absorption coefficient of ϵ280=14900 M−1·cm−1 were used to calculate protein concentrations. Samples of Bet v 1.0101 were prepared as described previously [26].

Size-exclusion chromatography

In order to determine the oligomerization state and the molecular mass of recombinant Fra a 1E, size-exclusion chromatography was performed with a Superdex S75 GL 10/300 column (total bed volume: 24 ml; GE Healthcare). Potassium phosphate buffer, 10 mM, pH 7.0, containing 150 mM NaCl was used as sample buffer as well as for column equilibration and for isocratic protein elution. Molecular mass calibration of the column was performed with albumin (67.0 kDa), ovalbumin (43.0 kDa), carbonic anhydrase (29.0 kDa) and cytochrome c (12.4 kDa).

CD spectroscopy

Far ultraviolet CD measurements were recorded on a Jasco J-810 spectropolarimeter (Japan Spectroscopic) equipped with a 1-mm quartz cuvette (Hellma). Spectra of 10 μM Fra a 1E and 7.0 μM Bet v 1.0101 were obtained at 20°C in 1 mM Hepes buffer, pH 7.0, containing 5 mM NaCl. At least 10 scans in the range between 260 and 190 nm with 50 nm/min scanning speed were averaged for each measurement, and the resulting spectrum was smoothed and normalized using the following formula to calculate the mean residue molar ellipticity: where Θ is the measured ellipticity, MRW is the mean residue mass, c is the protein concentration, d is the path length and N is the number of amino acids.

Computational methods

Multiple and pairwise sequence alignments were performed with ClustalW (EMBL-EBI) [27] and LALIGN (EMBL-EBI) [28], and homologous protein structures were aligned with the program Coot [29]. The cavity volumes of the Bet v 1 homologous proteins were calculated with CASTp with default parameters [30]. For NMR structures, the cavity volumes were determined for every single structure of the NMR bundle. The quoted volumes are means±S.D. Figures of protein structures were generated using the program PyMOL v1.0 (2009, DeLano Scientific LLC).

NMR spectroscopy

All NMR experiments were performed at 298 K on a Bruker Avance 800 MHz spectrometer with cryogenically cooled triple resonance probes equipped with pulsed field-gradient capabilities, and standard double- and triple-resonance through-bond correlation. NMR spectra to assign chemical shifts [31] were obtained with a 700 μM 15N, 13C-labelled Fra a 1E sample in 20 mM Hepes, pH 7.0, 150 mM NaCl and 10% 2H2O. Three-dimensional 13C and 15N edited NOESY (nuclear Overhauser enhancement spectroscopy) experiments (mixing times 120 ms) were recorded for derivation of distance restraints. NMR data were processed using in-house software and visualized with NMRView [32].

Structure calculation

NOESY cross peaks were classified according to their relative intensities and converted into distance restraints with upper limits of 3.0 Å (strong), 4.0 Å (medium), 5.0 Å (weak) and 6.0 Å (very weak) (where 1 Å=0.1 nm). For ambiguous distance restraints, the r−6 summation over all assigned possibilities defined the upper limit. Dihedral restraints were taken from analysis of chemical shifts by the TALOS software package [33]. Hydrogen bonds were included for backbone amide protons in regular secondary structure, if the amide proton did not show a water exchange cross peak in the 15N-edited NOESY spectrum. Structures were calculated with the program XPLOR-NIH 1.2.1 [34] using a three-step simulated annealing protocol with floating assignment of prochiral groups including a conformational database potential.

The 20 structures showing the lowest values of the target function excluding the database potential were further analysed with XPLOR-NIH 1.2.1 [34], MolMol [35] and PROCHECK 3.5.4 [36]. The structure coordinates were deposited in the protein data bank (pdb accession number 2LPX), and chemical shift assignments were deposited in the BioMagResBank (BMRB accession code 18281).

RESULTS AND DISCUSSION

General comparison, CD spectroscopy and size-exclusion chromatography

According to sequence alignment with PR10 proteins with known structure, Fra a 1E is most similar to the major cherry allergen Pru av 1 with a sequence identity of 79.4% [6] (Figure 1 and Table 1). The glycine-rich loop motif, residues 46–51, is highly conserved throughout the Bet v 1 superfamily.

Figure 1

Sequence alignment of Bet v 1 homologous proteins with known structure

Multiple sequence alignment of Bet v 1 homologous proteins as performed with ClustalW: Fra a 1E (Fragaria x ananassa, strawberry), Pru av 1 (Prunus avium, cherry), Bet v 1 (Betula verrucosa, birch), LIPR-10.2A (Lupinus luteus) and Gly m 4 (Glycine max, soya bean). Amino acids are marked with asterisks (identical), colons (conserved) and dots (semi-conserved). The highly conserved glycine-rich loop region is highlighted by a grey box. Conserved cavity residues are highlighted in yellow. Secondary structure elements are indicated as present in Fra a 1E (bottom line: β-sheets, blue; α-helices, green).

Table 1

Comparison of Bet v 1 homologous proteins

Protein	PDB code (method)	Sequence identity to Fra a 1E (%)	Sequence similarity to Fra a 1E (%)	Backbone rmsd values (Å)	Backbone rmsd values for secondary structures (Å)	Cavity size (Å3)	Reference
Fra a 1E	2LPX (NMR)	100	100	–	–	2468±440	Present study
Pru av 1.0101	1E09 (NMR)	79.4	95.0	1.82	1.04	1643±280	[6]
Bet v 1.0101	1BTV (NMR)	53.8	81.2	2.33	1.71	1633±152	[45]
	1BV1 (X-ray)	53.8	81.2	1.85	1.32	1527	[45]
Gly m 4.0101	2K7H (X-ray)	52.5	78.1	2.32	1.71	1621±242	[9]
LIPR10.2B	2QIM (X-ray)	51.9	80.6	1.24	1.93	2181.7	[41]
Bet v 1.1001	1FM4 (X-ray)	51.2	80.6	1.28	1.93	2359	[38]
LIPR10.2A	1XDF (X-ray)	50.9	78.6	1.42	1.80	329.3	[46]

Size-exclusion chromatography of Fra a 1E revealed a single peak corresponding to a molecular mass of 20.8 kDa, in good agreement with the theoretical mass of 19.0 kDa (Figure 2A). No additional peaks could be detected, indicating that under the present conditions Fra a 1E is monomeric.

Figure 2

Biophysical properties of Fra a 1E

(A) Size-exclusion chromatography of Fra a 1E with a Superdex S75 GL 10/300 column. The run was performed with 0.35 mg of Fra a 1E in 10 mM potassium phosphate buffer, pH 7.0 and 150 mM NaCl. The fit to the data obtained for the molecular masses of the standard proteins (inset, open circles) was used for the determination of the molecular mass of Fra a 1E (closed circle). (B) Far UV CD spectra of Fra a 1E (continuous line) and Bet v 1.0101 (dotted line) were acquired at 20°C with a bandwith of 1 nm, a sensitivity of 100 mdeg and a data density of 10 points/nm in a 1 mm quartz cuvette with 10 μM Fra a 1E in 1 mM Hepes buffer, pH 7.0, containing 5 mM NaCl, and 7 μM Bet v 1.0101 in identical buffer.

The CD spectrum of Fra a 1E displays characteristic features of a folded protein (Figure 2B), but differs from the Bet v 1 CD spectrum in several respects: the minima wavelengths (Fra a 1E: 217 nm; Bet v 1: 215 nm) as well as the maxima wavelengths (Fra a 1E: 194 nm; Bet v 1: 196 nm) are slightly different, the Bet v 1 CD 196 nm has lower amplitude in maximum as well as minimum ellipticity.

The NMR spectra of Fra a 1E show signal dispersion characteristics of a well-folded protein, and the application of double- and triple-resonance NMR experiments yielded nearly complete assignments of all 1H, 13C and 15N resonances (see Supplementary Figures S1 and S2 at http://www.bioscirep.org/bsr/032/bsr0320567add.htm). In total 1601 experimentally derived restraints were obtained for the structure calculation. The resulting ensemble of 20 structures shows good stereochemical properties and only low violation of experimental and geometrical restraints, it is well defined with an overall backbone rmsd (root mean square deviation) of 0.92 Å for residues Gly2–Asp154 and 0.54 Å for the regions with defined secondary structure only. More than 90% of the residues are located in the most favoured regions of the Ramachandran map (see Table 2).

Table 2

Statistics for structure calculations

NOE, nuclear Overhauser effect.

Experimental restraints	Value
Distance restraints
NOE	1270
Intraresidual	52
Sequential	287
Medium range	340
Long range	591
Hydrogen bonds	122
Dihedral restraints	209
Restraint violation
Average distance restraint violation	0.0027±0.001 Å
Maximum distance restraint violation	0.13 Å
Average dihedral restraint violation	0.15±0.07°
Maximum dihedral restraint violation	3.5°
Deviation from ideal geometry
Bond length	0.00053±0.00002 Å
Bond angle	0.11±0.004°
Co-ordinate precision*
Val2–Asp154
Backbone heavy atoms	0.91 Å
All heavy atoms	1.42 Å
Defined secondary structure†
Backbone heavy atoms	0.54 Å
All heavy atoms	1.06 Å
Ramachandran plot statistics‡	90.3%, 7.9%, 0.7%, 1.1%

*The precision of the coordinates is defined as the average atomic rmsd between the accepted simulated annealing structures and the corresponding mean structure calculated for the given sequence regions.

†The region of defined secondary structure is composed of Val2–Ser11; Ala16–Ile34; Lys40–Glu46; Gly53–Thr58; Tyr67–Asp76; His80–Glu88; Glu97–Ala108; Gly111–Thr123; Lys130–Lys153.

‡Ramachandran plot statistics are determined by PROCHECK and noted by most favoured/additionally allowed/generously allowed/disallowed.

Description of the structure

The solution structure (Figure 3A) of Fra a 1E consists of a seven-stranded antiparallel β-sheet (residues Tyr6–Ser12, Lys40–Glu46, Gly52–Thr58, Tyr67–Asp76, His80–Glu88, Glu97–Ala108 and Gly111–Thr123) that is closed by a long C-terminal (residues Glu131–Asp154) and a short V-shaped α-helix (residues Ala16–Ile34) with a kink at Asp26. The helices and the β-strands encompass a very large hydrophobic cavity, as is typical for Bet v 1 homologous proteins. A comparison of the secondary structure elements of Fra a 1E with the structures listed in Table 1 reveals backbone atomic rmsd values between 1.04 and 1.93 Å, with the most closely related structure being that of Pru av 1 [6] which also has the highest sequence homology to Fra a 1E. Whereas these proteins show the same overall fold, there detailed structures contain subtle differences.

Figure 3

The overall structure of Fra a 1E

(A) Cartoon representation of the lowest energy solution structure of Fra a 1E (α-helices, green; β-sheets, blue; loop-regions, grey). (B) Backbone overlay of the 20 lowest energy solution structures of Fra a 1E, with backbone rmsd values (0.21–1.58 Å, blue; 1.58–2.95 Å, dark green; 2.95–4.32 Å, light green; 4.32–5.69 Å, yellow; 5.69–7.06 Å, orange; 7.06–8.43 Å, pink). Higher rmsd values generally correspond to higher structural flexibility.

For example, strands β5 and β6 are bent away from the hydrophobic pocket (Figure 4, inset, grey arrow) in Fra a 1E, and the orientation of the C-terminal helix differs between the homologous proteins; both variations influence size and shape of the hydrophobic cavity.

Figure 4

Comparison of Fra a 1E with homologous proteins

Overlay of the average Fra a 1E structure (2LPX, pink) with the structures of Pru av 1 (1E09, brown), Bet v 1.0101 (1BV1, dark grey), the hypoallergenic Bet v 1 isoform Bet v 1.1001 (1FM4, light grey), Gly m 4 (2K7H, dark green), LIPR-10.2A (1XDF, orange) and LIPR-10.2B (2QIM, light green). The structures are displayed in cartoon representation. The insets show the glycine-rich loop region, residues Gly47–Thr53, in stick representation and the region around β-strands 5 and 6 in cartoon representation, colours as above.

Several studies were focused on ligand binding to this hydrophobic pocket. Binding, transport or storage are obvious possibilities for the physiological function of these proteins [6,37-40]. A range of ligands that might be of physiologic relevance could be identified in an ANS (8-anilinonaphthalene-1-sulfonic acid) displacement assay with Bet v 1.0101, and compounds such as flavonoids, fatty acids and cytokinins bind with low micromolar affinities to the protein [37]. Interestingly, the cytokinin kinetin seems to have a binding site alternative to the hydrophobic cavity. Structural studies revealed that two or three ligand molecules can bind inside the pocket, mainly stabilized by hydrophobic and a few hydrogen-bonding interactions [38,41,42].

The structural differences are also reflected in the differences of the CD spectra. Fra a 1E contains 32% α-helical, 39% β-sheet and 29% unstructured regions, while Bet v1 shows 24% α-helical, 39% β-sheet and 37% unstructured regions.

The hydrophobic cavity

Like in other family members, the cavity of Fra a 1E is accessible via three structural gaps, between the long C-terminal α-helix and β1, close to the glycine-rich loop, and between C-terminal α-helix and the loop from Thr58–Tyr67.

The first gap is not present in all homologues as its presence depends on exact shape and arrangement of the C-terminal helix. The second gap is quite small, and its size of approximately 4–9 Å depends on the nature of the amino acid side chains in different homologues (in Fra a 1E: 5 Å). The glycine-rich loop between β2 and β3, from Gly47–Thr53, has the consensus sequence GXGGXGT which is highly conserved throughout the Bet v 1 superfamily (rmsd of 0.35 Å for the alignment of all loop atoms, Figure 4, inset). This loop was suggested to be a binding region for kinetin [39] and to play an essential role in IgG and IgE binding [6,7,43,44], making it a general IgE binding epitope responsible for cross reactions between Bet v 1 homologous proteins [4].

The third gap, between the C-terminal α-helix and the loop from Thr58–Tyr67, is the largest. Compared with Bet v 1 and Pru av 1 [6,45], this loop is less well defined in Fra a 1E (Figure 3B). The loop flexibility is strongly related to the cavity volume, as can be judged from the volume difference of approximately 200 Å3 between the Fra a 1E structures with the highest rmsd difference of the loops.

The major difference between other members of the Bet v 1 superfamily and Fra a 1E is the large volume of its hydrophobic pocket of roughly 2500 Å3. The cavities of Bet v 1, Pru av 1 or Gly m 4 are only approximately 1600 Å, and the cavity of LIPR-10.2A is even smaller at 330 Å3, due to a strong inward kink of the C-terminal helix. The flexible orientation of the loop Thr58–Tyr67 as well as the displacements of strands β5 and β6 in Fra a 1E also contributes to the remarkable size of its cavity. Compared with the homologous structures, the main cavity entrance is much larger in Fra a 1E (Figures 5A and 5B) due to the different orientation of the C-terminal helix and the rearrangement of amino acids in the N-terminus of the helix. This helix aligns well with those of Pru av 1 and LIPR-10.2B (Figure 4) and, in addition, in Fra a 1E the helix residues Glu131–Glu139 are shifted away from the hydrophobic core, moving the larger amino acids such as Lys130, Glu131, His132, Lys135 and Glu139 away from the cavity entrance. The increased size of this gap is also due to the movement of loop Pro36–Ala38 away from the entrance on the opposite side of the C-terminal helix, so that the protein backbone and Gln37 are not restrictive.

Figure 5

The hydrophobic pocket

Stereoscopic view of surface representations of (A) Fra a 1E (green) and (B) Bet v 1 (red), with orientation close to Figure 3 displaying the large entrance to the hydrophobic cavity.

The residues inside the Fra a 1E cavity are very similar to those of homologous proteins, particularly Pru av 1 ([6], Figure 1). The hydrophobic residues Phe20, Tyr82, Tyr84, Ile99, Tyr101 and Tyr121 are widely conserved and adopt similar conformations in the different structures, pointing to similar functions. Indeed, these amino acids were suggested to establish van der Waals interactions with ligands [6,38,40].

Structural relatives

In spite of the overall folds being highly similar, the structural alignment of the homologous proteins reveals several differences, and, although Fra a 1E may well play a role in substrate binding, storage, or transport, the substrate class may be different from the other Bet v 1 proteins. LIPR-10.2A and LIPR-10.2B from yellow lupine are rather flexible proteins that can undergo conformational adaptations upon ligand binding. This might explain the large variations of the cavity volumes observed for LIPR-10.2A, LIPR-10.2B, Bet v 1.0101 and Bet v 1.1001 (Table 1). LIPR-10.2B binding of three trans-zeatin molecules inside the pocket leads to an increase of the cavity volume by 1800 Å3 [41,42,46] when compared with LIPR-10.2A in the empty state. The Bet v 1.1001 cavity volume increases by 700 Å3 when two molecules of deoxycholate are bound [38]. Hence, the 2500 Å3 cavity of Fra a 1E provides for accommodation of more copies of a single ligand or, alternatively, a larger ligand.

As Fra a 1E is expressed in different plant compartments, it may fulfil various functions [25]. In addition to all the ligand-binding studies it was observed that Bet v 1 is able to bind and permeabilize membranes, and ligand binding may facilitate translocation [47]. One may thus speculate that Fra a 1E is involved in the transport of hydrophobic ligands into plant vacuoles for storage, or that it may bind hydrophobic compounds necessary for cell growth, such as plant hormones or flavonoids. To contribute to ending these speculations, the current focus of our studies is the identification of physiological ligands of the Bet v 1 superfamily proteins.