Results:
Commonα-hairpin fold provides a structural basis for IgE cross-reactivity
Examining the A1LS sequence reveals a single VBP motif with an additional pair of cysteines at both N and C-terminals, which likely fold to form a single unit. In contrast, the JR2LS contains three VBP motifs (Figure 1A and 1B). The J2LS degrades into a series of 5-7 kDa fragments corresponding to the individual α-hairpins [16]. Based on this background, the three J2LS VBP motifs (J2.1, J2.2, J2.3) were produced recombinantly and their structures analyzed individually. Circular dichroism (CD) spectra of all four VBPs (A1LS, J2.1, J2.2, J2.3) show distinct minima at 222 and 210 nm consistent with predominantly α-helical secondary structure (S1). Further studies on the full-length J2LS using small-angle X-ray scattering (SAXS) yielded a radius of gyration of 25±2Å; significantly greater than the 20Å expected for a globular protein of equivalent molecular weight. Additionally, the shape of the corresponding Kratky plot suggested a high degree of flexibility. This data supports the proposed model in which J2.1, J2.2, and J2.3 could exist as independently folded hairpins separated by unstructured linker regions (S2), and validates our approach in examining each of these LSs as separate entities rather than part of a larger globular system.
The structural similarity afforded by the common α-hairpin scaffold could potentially give rise to cross-reactive epitopes despite the lack of sequence identity (Figure 1C). To explore this possibility, the structure of the WN and PN LSs was assessed using solution-NMR. Backbone and side-chain assignments were obtained using standard triple-resonance approaches. Analysis of the resulting chemical shifts using the TALOS prediction algorithm reveals two α-helical regions centered on the CxxxC motifs (S2). The downfield shift (>33 ppm) of the cysteine Cb peaks, along with the presence of NOESY cross-peaks between the Cb protons on opposite CxxxC repeats support the presence of an α-hairpin disulfide bonding pattern (S3). The A1LS contains an additional pair of cysteine residues located on either side of the main CxxxC motif (Figure 1B), which were found to form a third disulfide bond that was verified by mass-spectrometry (S1), explaining the high stability and survival of the intact form in the seed. This information, along with structural restraints derived from the available NMR data was used to determine the 3D structure of all four LS fragments. As shown in Figure 1A, all four constructs adopt an unambiguous α-hairpin structure with low backbone RMSD values (S3). Few long-range NOE interactions were detected beyond the disulfide bonded cysteines in each of the 4 LS fragments; a rather unusual observation for an ordered protein (S3). To further verify the calculated structures, rotational correlation times (Tc) were measured for all constructs (Figure 1A). The resulting values are consistent with a >7 kDa globular protein and not with an unfolded peptide. Taken together, this data suggests that A1LS, J2.1, J2.2 and J2.3 each adopt a common α-hairpin structure maintained by the conserved disulfide bonds. This architecture is common to the VBPs from other plant species [14, 27, 28], and provides a consistent scaffold that allows for permutations in amino acid sequence as shown by the low sequence identity beyond the conserved CxxxC motifs.
Independently of the IgE binding analysis, the ability of these α-hairpin structures to support cross-reactive epitopes was evaluated using the SPADE computational tool developed by Dall’Antonia et al. [29]. The A1LS amino acid sequence is shown (Figure 2A and 2B) with previously identified epitopes and the location of the overlapping peptides on the structure, respectively. The SPADE algorithm takes into account the physical properties of each residue similar to the PD metric employed for sequence analysis, but includes an additional layer using the 3D structure to examine solvent accessible area and spatial positioning from which regions of similarity corresponding to putative cross-reactive 3D epitopes can be identified. Comparing the structure of A1LS with the three J2LS VBP’s reveal areas of high surface similarity (ΔSIM) (Figure 2C), representing regions of potential cross-reactivity. However, this similarity was markedly reduced in J2.3 despite the conserved α-hairpin fold. The reverse comparison yields similar results (Figure 3A), with JR2.3 once again displaying noticeably lower overall (ΔSIM) values against A1LS than its stablemates. The amino acid sequence of the J2LS VBPs and the location of the overlapping peptides are shown in Figure 3B and 3C.