Determinants for dimer formation
For the study reported herein, the loop sequences of ELG and the secondary structure sequences of BLG were combined to create the chimeric LG, Gyuba. The Gyuba structure was characterized by X-ray crystallography, CD spectroscopy, gel filtration chromatography, and analytical ultracentrifugation, all of which indicated that Gyuba dimerized in a manner similar to that of BLG. Specifically, the RMSDs for the Cαs of residues between 5–60 and 67–152 (including those of the I-strand) of Gyuba and the different crystal structures of BLG are similar (Table II). Although the loops of Gyuba contain 30 residues that differ from those found in BLG, the substitutions seem to have very little effect on its secondary structure. On the other hand, the ability of Gyuba to dimerize was achieved by altering the ELG sequence at 40 positions. Because LPI is monomeric,21 it is apparent that, of those 40 residues, the nine substitutions surrounding the I-strand were not sufficient for dimerization. The residues responsible for dimer formation reside instead in the 31 residues of the other secondary structures. Although a high-resolution structure of ELG is unavailable, our observations suggest that at least some of the 31 substitutions change the overall arrangement of the secondary structures in ELG, thereby inhibiting dimerization. Recently, the importance of incorporation of water molecules and increase of flexibility (entropic contribution) for dimer formation of BLG was suggested from thermodynamic and structural studies.26, 27 The 31 differences between ELG and BLG amino acid sequences may also affect hydration or dynamic properties of the molecules.
Stability of Gyuba
Although both Gyuba and BLG dimerize, the stability of Gyuba is substantially less than that of BLG and ELG (Fig. 6). Acid denaturation was common to both Gyuba and ELG and was not observed for BLG. Generally, the pH-dependent change in protein stability can be explained by differences in the pKa values of the native and denatured states28 and may be associated with an Asp or Glu carboxyl in the ELG sequence (Asp9, Glu14, Glu34, Glu77, Glu81, and/or Asp88) that has an abnormally low pKa in the native structure. We predicted the pKa values of these residues using the Gyuba crystal structure and PROPKA 3.0,29 but we found them to be within the expected range for Asp and Glu, that is, 3.96–5.44. Thus, we could not identify a residue(s) responsible for acid-induced denaturation.
It would be worthwhile to mention the predicted pKa value of Glu89, which is located at the EF loop and implicated in the so-called Tanford transition, was very high (8.11). In the crystal structure of Gyuba, the side chain of Glu89 was buried like as the corresponding residue of BLG with the closed loop conformation. Although Asn88 of the EF loop of BLG was replaced with Asp in Gyuba, this substitution does not affect the conformational property of the EF loop.
Even at neutral pH, Gyuba was much less stable than BLG or ELG, possibly because of the substitutions at positions 121, because of the loop/secondary structure interactions, and/or because of the conformational change around Cys66–Cys160 disulfide in Gyuba. In previous report, it was revealed that a Cys121Ala mutation reduced the stability of BLG by 6.5 kJ/mol, although it improved the reversibility.25 Because Gyuba has the same substitution, it must be at least partly responsible for destabilization of Gyuba. However, influence of Cys121Ala mutation on the cooperativity (m value) seems to be less than that of Cys121Ser or Cys121Val mutation.25 An interaction between Pro126 (in the loop that connects the H-strand and α-helix) and Tyr20 (in the A-strand) is found in BLG, whereas a corresponding interaction is absent in Gyuba because residue 126 is Gln. That interactions between proline and aromatic residues contribute to protein stability have been suggested by studies on the HP domain30 and exendin-4.31 In addition, a statistical study32 showed that 45% of all prolines in the surveyed proteins interact with aromatic residues, which suggests that prolines facilitate tertiary interactions. Thus, it is likely that the lack of a Pro at position 126 in Gyuba destabilizes its structure relative to BLG. The atoms of the disulfide-bonded pair, Cys66–Cys160, could not be fit to the electron density of the Gyuba crystal, because it was ill defined. In additional, the location of Cys66–Cys160 residues in Gyuba must differ from that in BLG, because Met156 of Gyuba was found at the position corresponding to Cys160 of BLG. That no free thiol groups were present in Gyuba was confirmed by a 5,5′-dithiobis(2-nitrobenzoic acid) assay performed using guanidinium-HCl-denatured Gyuba (data not shown). Therefore, a potential structural polymorphism involving Cys66-Cys160 in Gyuba may be responsible for the observed smeared electron density and relative instability of its overall conformation. Although other residues may partially destabilize Gyuba, the loops must substantially contribute to the relative instability.
Notably, both the Tm and the cooperativity of the thermal denaturation curve were lower for Gyuba compared with BLG and ELG. The values of ΔH for our thermal denaturations of ELG (250 kJ/mol at 70°C) and BLG (230 kJ/mol at 74°C) are consistent with those calculated from data acquired under somewhat different conditions (270 kJ/mol at 70°C, pH = 4, and 2M urea for ELG and 290 kJ/mol at 74°C, pH = 2 for BLG).33, 34 However, ΔH for the thermal denaturation of Gyuba (120 kJ/mol at 65°C) is ∼ 50% of those values. Given that the structures of Gyuba and BLG are similar, it is unlikely that, on unfolding, the number of interactions broken (e.g., hydrogen bonds, van der Waals interactions) or the hydrophobic area exposed differs substantially between these proteins. The relatively small ΔH calculated for thermal denaturation of Gyuba may be the result of an incorrect assumption, that is, the two-state model is not an accurate description of unfolding if a metastable intermediate(s) exists during the transition. Conformational polymorphism involving Cys66-Cys160 might cause a non-two-state transition.
Finally, it is interesting to consider how evolutionary selective pressure affects protein-folding cooperativity. Baker and coworkers35 used a computationally designed protein, Top7, to address this point and concluded that highly cooperative folding is likely the result of natural selection. Our results support this conclusion. Although Gyuba assumes a BLG-like conformation and dimerizes with a similar Ka, the cooperativity of its thermal denaturation is much lower than that of the naturally occurring proteins, BLG and ELG.