Solution structure of the major fish allergen parvalbumin Sco j 1 derived from the Pacific mackerel

Although fish is an important part of the human diet, it is also a common source of food allergy. The major allergen in fish is parvalbumin, a well-conserved Ca2+-binding protein found in the white muscle of many fish species. Here, we studied the solution structure of the parvalbumin Sco j 1, derived from the Pacific mackerel, using nuclear magnetic resonance spectroscopy. We mapped the IgE-binding epitope proposed in a recent study onto the present structure. Interestingly, three of four residues, which were elucidated as key residues of the IgE-binding epitope, were exposed to solvent, whereas one residue faced the inside of the molecule. We expect that this solution structure can be used in future studies attempting to analyze the various IgE-binding modes of these allergens.

studies hypothesized that IgE can recognize the linear (i.e., sequential) epitope of Sco j 1 16 . Kubota et al. reported that heat treatment of Sco j 1 at 140 °C completely destroys its IgE reactivity 17 . In contrast, Kobayashi et al. showed that Ca 2+ binding of Sco j 1 maintains its IgE reactivity by using assays in the presence or absence of Ca 2+ chelating agents 18 . These studies showed that IgE recognizes the conformational (i.e., structural) epitope of Sco j 1. Given that both linear and conformational epitopes were found, the molecular basis underlying Sco j 1 reactivity to IgE is highly complicated and difficult to interpret. Thus, to analyze the details of recognition between Sco j 1 and IgE, further investigations are needed based on the precise three-dimensional structure of Sco j 1. However, the three-dimensional structure of Sco j 1 has not yet been reported; it has only been modeled using a computational method. This study aimed to determine the three-dimensional structure of Sco j 1 at atomic-scale resolution using NMR spectroscopy.

Results and Discussion
Solution structure of Pacific mackerel parvalbumin Sco j 1. All of the backbone amide resonances of Sco j 1 were assigned except for Asp80 (Fig. 1). Nearly complete side-chain assignments were also accomplished. The solution structure of Sco j 1 was calculated based on 2067 inter-proton distances and 173 dihedral angle restraints. Figure 2a shows the ensemble of the 20 superimposed lowest energy structures. For residues 2-109, the mean pairwise root mean square deviations were 0.44 Å for the backbone heavy atoms and 0.80 Å for the all heavy atoms, indicating that the solution structure of Sco j 1 was well defined. Statistics of the calculated structures are listed in Table 1, showing that they satisfy the NMR-derived restraints. Figure 2b shows a ribbon representation of the lowest energy structure of Sco j 1. Sco j 1 formed a single, compact domain consisting of six α-helices with residues 9-19 (helix-A), 27-34 (helix-B), 41-51 (helix-C), 61-66 (helix-D), 80-90 (helix-E), and 100-106 (helix-F). These secondary structural elements are in good agreement with those of the other parvalbumin structures 14,19 . Helix pairs AB, CD, and EF form three EF-hand motifs. The loop regions flanked by the helix pairs CD and EF are involved in calcium binding, whereas the N-terminal AB site is a non-functional EF-hand motif.
Comparison with other parvalbumin structures. For structural comparison, the solution structures of Sco j 1 and other parvalbumin isoforms from Atlantic cod 14 (Gad m 1; PDB ID: 2MBX) and carp 19 (Cyp c 1; PDB ID: 1CDP) were inputted into the DALI server (http://ekhidna.biocenter.helsinki.fi/dali_server/). The identity of amino acid sequences among fish parvalbumin is relatively high at 67-85%, suggesting that their three-dimensional structures are also similar. Sco j 1 showed high structural similarity with both Gad m 1 and Cyp c 1 ( Fig. 2c and Table 2). In particular, the Z-score of 18.8 between Sco j 1 and Cyp c 1 shows that they have similar global folding structures. However, the IgE-binding conformational epitopes of parvalbumin differ with regard to not only global folding but also the electrostatic distribution, solvent accessibility, and dynamic properties of each residue. Therefore, determining the three-dimensional structure at atomic resolution is important for further investigation of allergen proteins.
Epitope mapping. IgE-binding epitopes of parvalbumin have been suggested to be located in various regions. Using synthesized peptide fragments, it was previously found that the region from Ala22 to Thr41 contains the IgE-binding epitope in Sco j 1 16 . As shown in Fig. 3a, this region is located on helix-B and its surrounding region. Given that the peptide-based approach was used, this region is most likely a linear-type epitope. For Cyp c 1, the IgE-binding epitopes fall within some areas of the following residues (23, 25-29), (33-37), (77-79), and (87, 89-92, 94) 20 . The first two areas approximately overlap with the epitope of Sco j 1. Given that these epitopes were identified using the phage display technique, they are thought to be conformational epitopes. Surprisingly, for Gad m 1, the epitope is found at the C-terminal region of residues 102-109, which is quite different from the location for Sco j 1 6 .
In Fig. 3a, the sidechain atoms of Lys28, Lys29, Cys34, and Lys39 are represented by stick models. Yoshida et al. have shown that substitution of each residue into Ala reduces the reactivity by more than 50% in the ELISA experiments. Thus, these residues are elucidated to the key residues in the IgE-binding epitope. Whereas Lys28, Lys29, and Lys39, which are hydrophilic amino acids, are exposed to solvent, the sidechain of Cys34 faces the inside of the molecule. Thus, Cys34 is exposed to IgE binding after the molecule is decomposed in human digestive organs. Lys28 and Lys39 are fully conserved residues in Sco j1, Cyp c 1, and Gad m 1. Therefore, these residues are elucidated to play an essential role in IgE binding and crossreactivity.
As mentioned above, IgE recognizes Sco j 1 by its specific peptide region that acts as the sequential epitope. However, some evidence suggests that the IgE reactivity of Sco j 1 is maintained by its three-dimensional structure. First, substitutions of Asp51 and/or Asp90, i.e., residues consisting of Ca 2+ -binding CD and EF sites, respectively, to alanine significantly reduced IgE reactivity 21 . Similarly, the addition of ethylene glycol tetraacetic acid (EGTA) also reduces IgE reactivity 18 . These studies indicate that depletion of the Ca 2+ binding ability results in a conformational change of parvalbumin; thus, IgE must recognize Sco j 1 by some type of conformational epitope. Second, it was previously reported that the IgE reactivity of muscle extracts from Pacific mackerel gradually decreased as the temperature was increased to 120 °C and was completely lost at 140 °C 17 . This also indicates that IgE recognizes the three-dimensional structure of parvalbumin.  Table 2. Structural similarity of Sco j 1 to other parvalbumins.

CYANA noeassign manual restraints
In conclusion, this study investigated the solution structure of Sco j 1, a Pacific mackerel parvalbumin allergen, which was found to be similar to that of Gad m1 and Cyp c 1 derived from other fish species. The IgE-binding epitopes of Sco j 1 are located on one side of the molecular surface. Further investigation into the interaction mechanism between IgE and IgE-binding epitopes for both sequential and conformational recognition modes will be addressed using purified monoclonal antibodies for Sco j 1.

Methods
Protein expression and purification. The amino acid sequence of Sco j 1 was retrieved from the UniProt database (accession code P59747). The DNA encoding full-length Sco j 1 (1-109), optimized in the codon usage for Escherichia coli, was cloned into a pGEX-6P-1 expression vector. The actual synthesis and purification of the vector was outsourced (GenScript, Tokyo, Japan). The expression vector was transformed into BL21(DE3) competent cells. The cells were grown in 1 L MP media containing 2 g of 15 NH 4 Cl (Shoko Science, Tokyo, Japan) and 4 g of 13 C-labeled glucose (CIL, USA) at 37 °C. Protein expression was induced at OD 600 = 0.7 by adding isopropyl β-D-1-thiogalactopyranoside to a final concentration of 0.5 mM. The induced cells were cultured at 37 °C for 3 h. After harvest, disruption, and centrifugation, the expressed protein was purified using a glutathione Sepharose 4B column (GE Healthcare Life Science, USA) in a bed volume of 5 mL and eluted with 10 mM reduced glutathione-containing buffer. The GST tag was excised from Sco j 1 by incubation with HRV3C protease for 12 h at 4 °C. The 10 mL of isolated Sco j 1 was purified using an ÄKTA Purifier and a HiLoad 26/600 Superdex 75 gel-filtration column (GE Healthcare Life Science, USA). The protein was concentrated to 0.5 mM using Vivaspin 15 R (Sartorius, Germany) in NMR buffer containing 7% D 2 O, 20 mM MES, 150 mM NaCl, and 10 mM CaCl 2 at pH 6.8. NMR data collection, assignments, and structure calculation. NMR spectra were obtained at 25 °C using the Avance III HD 800 MHz (Bruker BioSpin, USA) and Unity INOVA 600 and 500 MHz (Agilent, USA) spectrometers. Two-and three-dimensional NMR spectra were processed using NMRPipe 22 , and the data analysis was performed using the Sparky program. 1 H-, 13 C-, and 15 N-resonances were assigned using the following set of spectra: [ 1 H- 15  Inter-proton distance restraints for structural calculation were obtained from 13 C-edited NOESY-HSQC and 15 N-edited NOESY-HSQC spectra, using 75 msec of mixing time. The restraints for backbone phi and psi torsion angles were predicted from chemical shifts of backbone atoms, using the TALOS + program 24 . The structure was calculated using the CYANA 2.1 software package 25 . For calculation with calcium ions we used the canonical EF-hand Ca 2+ -binding topology. Sco j 1 has two typical EF-hands regions: D 52 QDKSGFIEEEE 63 and D 91 SDGDGKIGIDE 102 . Two calcium ions were coordinated one by one at each binding site. The upper limit distance from the atomic group involved in coordination was 2.8 Å.
NOE distance constraints were automatically assigned and calculated using seven cycles under the "noeassign" macro of CYANA. To correct for the automatic NOE signal assignment, we used additional distance restraints derived from manual NOE assignments. Based on a model structure of parvalbumin, constructed using MODELLAR 26 from a carp parvalbumin structure (PDB ID: 1CDP), all NOE signals were manually assigned. Among these assigned signals, we extracted the restraints between the medium-and long-range residues, and used a pseudo-atom for all optical isomeric protons. Consequently, we used 889 distance restraints with a fixed upper limit of 6 Å. At each "noeassign" cycle, 100 structures were calculated using 30,000 steps of simulated annealing, and a final ensemble of 20 structures was selected based on the CYANA target function values.
Data availability. BMRB accession number. The resonance assignments have been deposited to BMRB (code: 36086).
PDB accession number. The atomic coordinates have been deposited to PDB (code: 5XND).