植物が有するPhytochrome B(phyB、分子量130k)は、開環テトラピロールを発色団に持つ赤色光・近赤外光受容蛋白質である。phyBは暗中で赤色光吸収型(Pr型)にあるが、赤色光によって近赤外光吸収型(Pfr型)に光可逆的に変化する。Pfr型は細胞核内に移行して遺伝子制御蛋白質と相互作用し、様々な遺伝子の発現制御を通じて植物個体の光形態形成応答を誘起する。本研究では、電子顕微鏡などにより、Pr型の立体構造を明らかにすることを目指した。
構築済み発現系によって、シロイヌナズナ由来phyBの高純度標品を得、放射光X線小角散乱(SAXS)実験を実施した。phyBが二量体として存在することを明らかにするとともに、分子概形推定法を用いてPr型の分子概形を決定した。また、Pfr型で光可逆的凝集が生じることも明らかにした(Oide et al. (2019) FEBS J.)。
負染色Pr型の電子顕微鏡像分類から、SAXSモデルと大きさや形状が酷似した像が得られた。さらに、透過型クライオ電子顕微鏡を用いて急速凍結したPr型の分子像を観察した。扁平形状のphyBは、同程度の分子量を持つ球状蛋白質に比べてコントラストが低かったが、SAXSモデルと矛盾しない平均分類像を得た。しかし、0.1秒以内にphyB分子が凍結前の水薄膜表面にブラウン運動で到達し、選択配向と分子変性が生じている可能性が高く、三次元再構成像の分解能は0.8 nmにとどまっている。結晶化も含めて、気液界面での選択配向や変性を回避する方法を模索した。
phyBは、構造揺らぎが大きいと予想され、構造多形を考慮した解析方法が不可欠である。ドメイン運動が顕著な標準試料グルタミン酸脱水素酵素についてクライオ電子顕微鏡解析を行い、構造多形の近原子分解能三次元構造を明らかにし、ドメイン運動の自由エネルギー地形を推定法を考案した(Oide et al. (2020) FEBS J.)。また、電子顕微鏡では水和水分子の可視化が困難なことから、将来の構造解析に備え、人工知能を用いた水和構造予測プログラムの開発を試みた。
Plant phytochrome B (phyB) is a red/far-red-light receptor protein with phytochromobiline as a chromophore. PhyB is in red-light-absorbing (Pr) form in the dark. When irradiated by red light, phyB is converted to far-red-light absorbing (Pfr) form to control gene expression in nuclei. The activity of phyB in Pfr form induces various types of photomorphogenic responses in plants. In this study, the structure of Pr form was studied by using cryo-electron microscopy (cryoEM) and small-angle X-ray scattering (SAXS).
PhyB from Arabidopsis was expressed and purified by using a protocol established in our laboratory. The molecular organization and shape of Pr in solution were studied by SAXS using synchrotron X-rays at SPring-8. The SAXS profile revealed that phyB exists as a dimeric form. In addition, by applying an ab initio molecular shape determination algorithm with a protocol for selecting most probable molecular shape, the molecular model of Pr at a low resolution was determined. In addition, SAXS experiment revealed an aggregation property of Pfr form (Oide et al. (2019) FEBS J.).
The molecular shape of phyB was also studied by electron microscopy for negatively stained Pr. After two-dimensional classification, we obtained molecular images very similar to the SAXS model. CryoEM observations toward structure analysis at near-atomic resolution were carried out by using the latest model of cryo-electron microscopy. Purified phyB solution was irradiated by far-red light to accumulate Pr form before flash-cooling. Since phyB has a planar molecular shape, the contrast in electron miscopy images was lower than those of globular proteins with molecular weight of 280 K. Classification of picked up molecular images gave a set of averaged shape of Pr explained as different views of the SAXS model. However, probably because of Brownian motion of phyB in thin liquid film of 100-nm thickness, phyB molecules reach to the liquid-air interface within 0.1 s much longer than the time necessary to flash-cooling procedure. At the interface, phyB molecules are in preferred orientation and/or denaturation. Therefore, nominal resolution of a reconstructed three-dimensional electrostatic potential map is limited at 0.8 nm. To improve the resolution as well as to try crystallization, we are searching method necessary to avoid preferred orientation and denaturation.
Since phyB molecule is composed of six functional domains connected with loops, any structural analysis protocol incorporating structural heterogeneity in solution is necessary. In this study, we developed an analysis method to visualize metastable conformation and free-energy landscape of them from electron microscopy images. The method was applied to a standard specimen, glutamate dehydrogenase, and we successfully visualized four metastable conformations at near-atomic resolution with their free-energy differences(Oide et al. (2020) FEBS J.). In addition, electron miscopy is difficult to identify oxygen atoms of proteins and also water molecules in hydration layer. To propose molecular model of phyB including hydration layer, we started the development of software to predict hydration structure of any proteins by using deep learning.
|