Structural basis for cross-reactivity and conformation fluctuation of the major beech pollen allergen Fag s 1

Fag s 1 is a member of the Pathogen Related protein family 10 (PR-10) and can elicit cross-reaction with IgE antibodies produced against the birch pollen allergen Bet v 1. The Nuclear Magnetic Resonance (NMR) structure of Fag s 1 is presented along with its dynamic properties. It shares 66% identity with Bet v 1 and exhibits the expected three α-helices and seven β-sheets arranged as a semi-beta barrel and exposing the residues mapped as the Bet v 1 IgE epitope. The structural dynamics of Fag s 1 were monitored on the fast and intermediate timescales, using relaxation rates. The complex dynamics of Fag s 1 are closely related to the internal cavity, and they modulate IgE and ligand binding.

quantitative analysis performed by Grutsch et al. (2014) confirmed that ligand binding to the interior cavity leads to a compaction of the three-dimensional structure accompanied by a decrease in chemical exchange in the protein backbone and formation of a less dynamic conformer 11 .
Bet v 1 assumes various conformations in the free state that might represent the cavity opening, thus permitting the ligand to enter and bind to it. This conformational exchange process could also be important for other proteins with similar folds Therefore, characterizing the mechanisms of these transitions at the atomic level can significantly impact our understanding of the biological functions of many PR-10 proteins and shed light on the differences in allergenicity observed among isoforms, mutants, and related proteins.
In this work, we determined the NMR solution structure and the structural dynamics of Fag s 1 by monitoring on the fast and intermediate timescales. This work shows that the conformational exchange around the inner cavity of Fag s 1 is responsible for its complex dynamics.

Results
NMR solution structure of Fag s 1. The solution structure of Fag s 1 was calculated using NMR restraints, and it is composed of three α-helices and seven β-strands arranged as a curved semi-beta barrel. Figure 1a shows the lowest-energy structure as a ribbon representation, and Fig. 1b shows an ensemble of the ten lowest-energy structures. Three α-helices, α1, α2, and α3 are formed by residues P15-F22, A26-V33, and E130-E152, respectively. Residues V2-T11, I38-I43, I53-F58, Y66-D75, T80-G88, L95-A106, and S112-T122, are organized in β-strands β1-β7. The structure calculation statistics are summarized in Supplementary Table 1 Figure 1c shows the superposition of Fag s 1 (PDB ID: 6ALK) on Bet v 1 (PDB ID: 4A88) 4 . The two structures are similar, showing small differences in α1 and in the C-terminal of α3 (Fig. 1c). Figure 1d,e show the Fag s 1 structure in a ribbon representation and its surface, highlighting the antibody epitope mapped in Bet v 1 using NMR and X-ray. Most of the mapped epitope is conserved in the Fag s 1 surface except for N65, I71, N76, A77, T117 and K145. Supplemental Fig. 1 shows the primary sequence alignment of several Bet v 1 homologs including Fag s 1, demonstrating that the epitope is conserved and present in all the sequences.
Supplementary Table 2 summarizes the known structural information for Bet v 1 homologs compared with Fag s 1. All show strong similarities in their primary sequence (up to 71%) and tertiary structures. The Dali score between Fag s 1 and Bet v 1 is 19.3 2,3,12 The cavity volume of Bet v 1 and its homologs is between 2,000 and 3,000 Å 3 (Supplemental Table 2). None of those features has a direct correlation with hypersensibility.
Backbone molecular dynamics of Fag s 1 by NMR spectroscopy. The molecular dynamics of Fag s 1 in the ps-ns timescale were monitored by examining the relaxation rates (R 1 and R 2 ) and the 1 H-15 N Heteronuclear NOE (Het-NOE) (Figs S2a,b and 2a). It was found to be monomeric (τ c 8.2 ns), rigid, having few residues -R70, E96, G111, I127 and K129 -with 1 H-15 N Het-NOE values below 0.65 (Fig. 2a).
The amide proton temperature coefficient has been used to predict hydrogen bond donors, and values more positive than −5 ppb/K are indicators of intramolecular hydrogen bonding 13 . The thermal coefficient index (TCI-ppb/K) of Fag s 1 was measured (Fig. 2b), and most of the NHs exhibited TCI values greater than −5 ppb/K, suggesting the presence of intramolecular hydrogen bonds, which it is consistent with the secondary structure elements except for E8, L18, S41, I74, A106, A135 and G136. The TCI values for the loop residues V12, L24, I38, G61, and E127 of Fag s 1 were less than −5 ppb/K. These results are consistent with a well-structured protein.

Conformational fluctuation of Fag s 1 as determined by relaxation dispersion experiments.
The molecular dynamics of Fag s 1 in the μs-ms timescales were monitored experimentally by relaxation-compensated 15 N-CPMG experiments (RC-15 N-CPMG) 14 . Figure 3a shows the values obtained for R 2,eff (s −1 ) when ν CPMG of 66.7 and 1000 Hz and the difference between them (ΔR 2,eff (s −1 )) as a function of the residue number of Fag s 1. A difference greater than 5 Hz was used to identify residues undergoing conformational exchange in the fast-to-intermediate regime on the NMR chemical shift timescale. Supplementary Figure 3 and 4 show the relaxation dispersion curves for selected residues.
In Fag s 1 and Bet v 1 some residues were identified as undergoing conformational exchange and side chains were found to point toward the cavity (Fig. 3b). The relaxation dispersion profile of four residues (F22, F58, F64 and L128) could not be evaluated because they showed broadened NMR signals and small signal to noise ratio, an indication of exchange. Figure 3c shows a detailed view of the of Fag s 1 cavity. The side chains of residues in conformational exchange are oriented in "bottlenecks" in the cavity, suggesting a correlation between movements on the μs-ms timescale experienced by Fag s 1 residues and fluctuations in the cavity shape and volume. For instance, the hydrophobic side chains of residues F22 in α1, L23 in L2, and I102 in β6 form "bottleneck 1" of the Fag s 1 cavity. Two phenylalanines, F58 in β5, and F64 in L5, form "bottleneck 2, " and the side chains of residues I133 in the N-terminal of helix α3 and I128 in L9 form the nearby "bottleneck 3. " Quantitative analysis of the conformational exchange of Fag s 1 was carried out by adjusting relaxation dispersion data using a theoretical global model of conformational exchange between two states: the ground state and the excited state. Relaxation dispersion curves were measured at 298, 300.5, 303, 305.5 and 308 K. From the theoretical treatment of relaxation dispersion curves acquired at five different temperatures, kinetic parameters could be obtained: exchange constant (k ex ) between the ground state (A) and the high-energy conformation state (B); the populations of each state (p A and p B ), and their difference in the 15 N chemical shift Δϖ 15N . From these  Table 3), Van't Hoff and Eyring plots were constructed, providing the energetic characterizations of high-energy state and transition states, respectively.
Data show that Fag s 1 residues are fluctuating at a k ex rate of 812 ± 60 s −1 and p b 8.5% at 298 K. Kinetic and thermodynamic analyses of the conformational exchange process experienced by Fag s 1 were performed using the reaction coordinate diagram shown in Fig. 3d. This diagram was constructed using the Gibbs free energies, enthalpies and entropies of the major state, transition state and high-energy state, with ΔG BA = 6 kJ/mol, at 298 K. The higher-energy conformation state is entropically unfavorable (TΔS = −43 kJ/ mol) and enthalpically favorable (ΔH = −37 kJ/mol).

Correlation between conformational exchange and interaction of Fag s 1 with dehydroergosterol (DHE).
Bet v 1 promiscuously binds molecules that include fatty acids, flavonoids, and cytokinins 5 as well as its putative physiological ligand quercetin-3-O-sophoroside 8 . Using the approach of Mogensen et al. (2002), that is an Anilo-8-Naphthalene Sulfonate (ANS) displacement assay, we evaluated the binding capacity of Fag s 1 with known Bet v 1 ligands such as Deoxycholate (DOC), Sodium Dodecyl Sulfate (SDS), quercetin, and dihydroergosterol (DHE) 6 . Fag s 1 interacted with all tested compounds with dissociation constants in the micromolar range, similar to that found Bet v 1, with the exception of quercetin, which interacted with much lower affinity (Fig. S4).
The binding of Fag s 1 with DHE was monitored using NMR spectroscopy in titration experiments. The chemical shift perturbation (CSP) index for the titration assay of Fag s 1 with DHE is shown in Fig. 4a. The NMR peaks of some residues at various concentrations of DHE are shown in Fig. 4b, and residues with higher CSP indices mapped onto the Fag s 1 solution structure (Fig. 4c). The NMR peaks of some residues disappeared during titration (in yellow in Fig. 4c).
Several residues involved in DHE binding also underwent conformational exchange on the μs-ms timescale, as mapped by relaxation dispersion experiments. To better understand this behavior, 15 N chemical shift variations (Δω 15N ), obtained by adjusting the DHE titration curves, were compared with those obtained by theoretical analysis of the relaxation dispersion experiments (Fig. 5a). It is worth noting that values obtained by titration and relaxation dispersion experiments were similar, except for residues 89 and 127 in β6 and β7. This result suggests structural similarity between the high-energy conformational state and the DHE-bound state.

Discussion
As with other proteins in the PR-10 family, the Fag s 1 NMR solution structure is composed of seven β-strands and three α-helices forming a secondary structure arrangement (β-α2-β6-α) with an antiparallel curved β-sheet. Those secondary structures fold into one tridimensional structure in which the main characteristic is the presence of a large hydrophobic or amphipathic cavity. This cavity plays a pivotal role in the function of all PR-10 proteins in ligand and IgE binding 8-12,15-17 . To illustrate this, we aligned Fag s 1 with Bet v 1 and eight other allergenic PR-10 proteins whose structures have already been determined, focusing on the Bet v 1-antibody epitopes. Mirza et al. solved the structure of the Bet v 1-Fab BV16 complex and from this structure it was possible to identify one conformational epitope composed of residues E42-T52 in the PR-10 highly-conserved p-loop, R70 and D72 in β4, H76 in loop L4, and K97 in β6 17 . Another important epitope-identifying study was carried out by Asam et al. (2014), who used polyclonal antibodies from Bet v 1-sensitive patients to identify conformational epitopes of Bet v 1 by NMR spectroscopy 10 . The location and composition of the Bet v 1-identified epitopes on Fag s 1 and Bet v 1 structures are conserved, except for residues N65, I73, A76, T117 and K145 (Fig. S1). The primary consequence of the great similarity between Fag s 1 and Bet v 1 is the cross-recognition of Fag s 1 by antibodies induced by Bet v 1 sensitization.
In contrast to its rigidity on the ps-ns time-scale, the Bet v 1structure is very dynamic at the μs-ms time-scale 10,11 . Movements at this time-scale are related to chemical conformational exchange, which contributes to the structural malleability of proteins, an essential characteristic for their biological functions. Usually, in order to perform their functions, proteins exist in different conformational states that can be representing intermediates of a folding process, mimicking of a bound state, or intermediates in a catalytic pathway [18][19][20][21] . Uniquely, NMR spectroscopy is a technique that allows characterization of μs-ms timescale movements in proteins and, consequently, provides important information about higher-energy conformational states [22][23][24] .
The molecular dynamics of Fag s 1 and its conformational equilibrium were characterized by relaxation-compensated 15 N relaxation dispersion experiments 14,20,25 . Many residues of Fag s 1 showed conformational exchange contributions in their R 2,eff (s −1 ) rates revealing that Fag s 1 assumes a ground state and a high-energy state in the apo form, as evidenced by the conformational dynamics on a μs-ms timescale.
Many of the identified residues have side chains located in the Fag s 1 cavity. Because many residues showing conformation exchange in the relaxation experiments were "strategically" positioned in bottlenecks in the Fag Residues in green showed CSP higher than the mean plus one standard deviation. Black dots represent residues whose NH peaks does not appear in the HSQC even in the free state and blue peaks with signal overlaps (b) Spectra of selected residues in different DHE concentrations. (c) Residues mapped in (a) are colored green in the Fag s 1 structure and in yellow those that disappeared during titration (I38, I56, I91, L104, S118, Y120, and I127).
SCieNtiFiC RePoRtS | (2018) 8:10512 | DOI:10.1038/s41598-018-28358-1 s 1 cavity, the idea of correlation between atomic movements on the μs-ms timescale as detected by NMR and fluctuations in the cavity shape and /or volume was reinforced.
Fluctuations in the internal cavity volume and/or shape seem to be the structural elements responsible for conformational variability. The "opening" and "closing" of the cavity could be responsible for formation of the higher energy state. Such transient conformational rearrangements could be necessary to generate a channel through which the ligand can enter the cavity, and they would include movement of the backbone, rearrangement of the loops and movement of the side chains.
Internal cavities in proteins are crucial structural elements that influence their dynamics and stabilities. For example, Phage T4 lysozyme is a well-studied protein, and its mutant, L99A, containing a large internal cavity capable of binding hydrophobic molecules, is one of the best-characterized systems. The creation of the cavity did not involve a change in structure but is related to protein dynamics, mainly involving residues close to the cavity 26 . In this case, the greatest difference between the ground and excited states is the F114 side chain, which is solvent-exposed in the major state and occupies the cavity created by the L99A mutation, becoming buried in the core.
Proteins face major challenges during the binding of hydrophobic molecules. They need to expose hydrophobic regions to permit entrance by the ligand, but as a result they can become vulnerable due to sampling of conformational states that present a risk of misfolding and aggregation [26][27][28] . Evolutionary pressure creates competition between increased binding efficiency and reduced protein stability, balancing the ability to breathe and populate multiple states without creating an unacceptable propensity for aggregation 29 .
Furthermore, conformation fluctuation close to the epitope-antibody binding site can modulate IgE affinity. We attempted to correlate the residues that exhibit conformation fluctuation with those that exhibit CSP in the presence of IgE, as reported by Asam et al. Fig. 5b,c show that these residues populate similar regions. The presence of residues in conformational exchange close to the binding site is a very common feature that is important for modulating affinity. Recently Yanaka et al. (2017), proposed a method called fluctuation editing, which identifies potential mutation sites using relaxation dispersion experiments 30 . Using this strategy, they were able to engineer a higher affinity antibody. Our data show potential sites for mutation that can help engineer a different allergen with special IgE binding properties.

Materials and Methods
Expression and Purification of Isotope-labeled ( 15  NMR Spectroscopy. The NMR experiments were performed at 308 K using Bruker Avance III 800 and 700 spectrometers equipped with TXI 5 mm triple-resonance probe and a Bruker Avance III 600 equipped with TCI 5 mm triple resonance cryogenic-probe. The assignment was done following procedures previously published by Moraes et al. 31 . For the assignment of the NMR resonances of the backbone atoms the following NMR triple-resonance experiment were analyzed: 1 H- 15 Table S1 were calculated using software Mole 2.0 36 and the webserver 3V 37 . The algorithm of Mole 2.0 is based on the computation of the Delaunay triangulation/Voronoi diagram of the atomic centers and further construction of the molecular surface, and identification of cavities. Cavities identified by Mole 2.0 should be analyzed carefully since they are geometric entities used to tunnel and pore identification. Differently, 3V webserver uses what is known as rolling-probe strategy, where the cavity is identified by taking the difference between the volume obtained by rolling on small probe sized similarly with the solvent radios and another probe with radius as large as possible. Both softwares have as refine parameters the size of both probes. In this work, we standardized smaller probe and bigger probe with radius of 1.3 Å and 3.0 Å, respectively. Backbone dynamics. The backbone molecular dynamics of Fag s 1 were characterized using the relaxation rates R 1 , R 2 and { 1 H-15 N}-Heteronuclear NOE (Het-NOE) obtained from 15 N relaxation experiments acquired in Bruker Avance III 500 MHz spectrometer. R 1 and R 2 rates were obtained from the fitting as a single intensity exponential decay of resonance intensities as a function of the delays: 0.05, 0.1, 0.2, 0.3, 0.5, 0.6, 0.8, 0.9 and 1.0 s for R 1   ) were obtained from the intensities measured using the software CCPN analysis 32 from 1 H-15 N correlation spectra and applying the following formula: where T relax is the relaxation delay of 30 ms, I CPMG is the intensity measured in 1 H-15 N spectra acquired with a CPMG frequency, ν CPMG , ranging from 66.7 to 1000 Hz. I 0 is the peak intensity obtained with T relax of 0 ms. where σ and I CPMG are the standard deviations of noise, and NMR resonance intensity measured from spectrum acquired with a ν CPMG of 500 Hz. A minimum error of 3% of the R 2eff (s −1 ) values where used when the experimental errors calculated using equation 2 where lower than this imposed value. The experimental R 2,eff (s −1 ) values as a function of ν cpmg were analyzed adjusting a theoretical model built up applying the numerical solution of Bloch-McConnell equations to a system in equilibrium between two states: a high-populated A state, ground state, and a low-populated B state, excited state 40 . The software CPMG_fit (Korzhnev and Kay, 2008) was used for fitting the dispersion curves and for extraction of kinetic and thermodynamic parameters. In this step, residues presenting overlaps on the 1 H-15 N HSQC spectra, those showing experimental results with low quality, and those in which the difference on R 2eff (s −1 ) obtained with ν CPMG of 66.7 (lowest value) and 1000 (highest value) was smaller than 5 Hz were excluded.
At the end, 15 N-RC-CPMG relaxation rates of n R = 22 amide groups, acquired at n T = 5 different temperatures and at n B = 2 magnetic fields were least-square fitted together by minimizing the following target function: are the experimental uncertainty associated with them. The summation in equation (3) was done over the number of experimental points ndat = {n R , n T , n B }. The data fitting was carried out taking the following assumption: (a) all residues in Fag s 1 are involved in the same global exchange process (same equilibrium rate constant, K ex ), (b) the intrinsic relaxation rates R 2,0 are the same in states A and B, (c) the chemical shift differences between states for each residue, Δϖ BA , are independent of temperature, and (iv) CPMG experimental data were analyzed independently for each temperature. The following adjustable parameters were obtained: population of state B, p B , and, consequently, the equilibrium rate constant constant, K e = p B /p A , rate constant k AB , n R 15 N chemical shift difference between states A and B, Δϖ BA, and n R *n B intrinsic relaxation rates, R 2,0 .
The Gibbs free energy difference of the states A and B, ΔG BA = ΔH BA -TΔS BA , was obtained fitting the temperature dependence of the equilibrium constant, K ex , using the Vant' off equation: where R = 8.314 J K −1 mol −1 is the gas constant, ΔS BA and ΔH BA are the entropy and enthalpy difference between states A and B, respectively, and T is the experimental temperature in K. The transition state was analyzed using Transition State Theory. The temperature dependence of rate constant k AB was fitted with the Eyring equation. is the chemical shift difference between the free and the DHE-bound state of Fag s, here and in the following sections of this work called ω Δ N 15 , for simplicity. The experiments were acquired in a spectrometer Bruker 600 MHz Avance III equipped with TCI 5 mm triple resonance cryogenic-probe at 308 K. Changes in dynamics of Fag s 1 caused by DHE binding either in μs-ms were monitored by the same experiments used to track the molecular dynamics of free Fag s 1. Following these ideas our work contributes to the understanding of the correlation between dynamics, ligand binding, allergenicity and IgE recognition of PR-10 proteins.