血清素受體蛋白亞型之動力模擬暨配體結合模式評估

Abstract

血清素受體均為穿膜蛋白，主要分佈在神經細胞上，分為七科，十四種亞型。除第三種亞型以外均隸屬 G 蛋白偶合受體（G Protein-coupled receptor, GPCR）。其訊息傳遞路徑在生理上具調控情緒、幻覺、記憶、橫紋肌收縮與飽食感等功能，也成為許多藥物分子的作用目標，臨床上已有抗精神病、抗憂鬱與偏頭痛藥物等應用。但多數血清素受體藥物以高度不具專一性著稱，副作用無法避免。而開發新藥時除了須讓藥物有能力辨識血清素受體與其他種類的 GPCRs ，受體亞型的選擇性（subtype selectivity）更是個尚待解決的議題。

不同受體與配體之間交互作用的差異，最好能由結構生物學探討其分子機制，即基於結構設計藥物（structure-based drug design）。然而，穿膜蛋白厭水與結構易變動的特性使其結晶困難，此類蛋白結構的解析仍為一大挑戰。目前實驗上尚未解得任何一種血清素受體結構，所幸近年陸續解出的數種 GPCRs 提供了電腦模擬的模板。在本論文中，將現存人類 GPCR 資料庫中 A 族群的所有序列與可得的結構模板，經多重序列比對結果引導，以現有視紫質蛋白（rhodopsin）與 β2 腎上腺素受體（β2-adrenergic receptor）為主要模版骨架，建構了 2A、2B、2C 與 4 至 7 等七種血清素受體亞型，以及補全缺失片段後的 β2 腎上腺素受體（作為控制組）之同源蛋白模型（homology/de novo models）。將這些初始蛋白模型埋入脂雙層膜--水系統中，運行奈秒時間尺度的分子動力模擬，探索穿膜蛋白在生理狀態下的未活化態結構。配體分子挑選實驗上具代表意義的選擇性抑制藥物，以量子化學執行結構最佳化後，參考模板結構，運用分子嵌合技術（molecular docking）預測藥物嵌入血清素受體的結合位置。 本論文挑選 5-HT7 與 5-HT2A 兩種受體亞型作為研究主體。在藥物的結合模式評估上，進行藥物受體複合物的分子動力學模擬，直接分析抑制性藥物在蛋白質受體內的運動結果。藉分析胺基酸與小分子官能基之間的交互作用差異，蛋白結合位置的大小與藥物分子佔據體積，可望在亞型選擇性的問題上做出分子層級上的解釋。此套模型建立的流程已半自動化，可望發展為一種相較於簡單評分函數更具物理意義的平台。日後可繼續拓展至其他血清素受體亞型，且進一步衍伸至親緣性相近的其他 GPCR 蛋白，如多巴胺受體（dopamine receptors）與鴉片受體（opioid receptors）等重要的藥物標靶之上。

An ideal serotonergic agent should not only distinguish its eponymous counterparts from other GPCRs, but also discriminate between subtypes. To achieve higher selectivity it is desirable to take a structure-based approach, and constructing in silico models by utilizing other GPCR structures is currently the only option for serotonin receptors. In this work, multiple alignment was conducted across class A human GPCR sequences for guiding homology (de novo) model constructions. Using the crystal structures of rhodopsin and β2-adrenergic receptor being major templates, seven distinct serotonin receptor subtypes 5-HT2A~5-HT2C and 5-HT4~5-HT7) were built. Selective 5-HT receptor inhibitors were chosen, structurally optimized by quantum-chemical calculations, and docked into the binding pockets. Each apo-and bound-form receptor was then immersed into a lipid bilayer/water environment and subjected to nanosecond-scale molecular dynamics (MD) simulations. Based on both induced-fit and conformational selection hypotheses, favorable ligand binding mode(s) were sampled either by ligand-bound form MD simulations, which mimicked the induced-fit process upon binding, or by relaxed complex scheme (RCS), which captured various conformers from MD snapshots to be 'selected' by massive dockings. The poses were then evaluated and ranked by scoring functions. Instead of exploring through all the subtypes exhaustively, the objective was first aimed on elucidating the molecular mechanism contributing to selectivity between 5-HT2A and 5-HT7 subtypes. By analyzing the 'molecular switches' of each receptor, the orientation of key residues, and the occupied volumes by drugs, several key components confering selectivity were found. The aforementioned process is applicable to other serotonin subtypes as well as non-serotonergic GPCRs, such as dopamine receptors and opoid receptors. Automated implementation is also a feasible option, and can provide a virtual screening platform which accounts for the flexibility on both ligands and receptors.