開發數據非依賴性採集質譜技術應用於大規模人類蛋白體定量

Abstract

利用串聯式質譜儀進行大規模蛋白體分析已經成為我們在探究細胞、組織或是其他臨床樣品中複雜的生物途徑時不可或缺的工具。然而，深入的蛋白質組定量通常需要在前處理過程中先將不同特性的胜肽分群(fractionation)、大量的樣品起始量以及耗時的儀器分析時間。這對於有限的臨床樣品造成了很大的限制。目前主流應用在大規模定量分析的數據依賴性採集技術(DDA)其伴隨的半隨機(stochastic)採集離子的特性，通常會導致在重複性實驗中會有較差的再現性。近年來，許多研究團隊開發了新的數據採集技術，數據非依賴性採集（DIA），這個方法在質譜中針對所有離子進行全面性的採集，並且與預先建立好的胜肽圖譜數據庫進行比對。此方法結合DDA的全面性的探索能力與目標選擇反應監測（MRM）的定量準確性。然而，建造一個高品質的圖譜數據庫並且深度涵蓋整體基因組的目標仍需要進一步開發及與改善。 為了透過DIA技術改善蛋白體分析覆蓋範圍，其中幾個關鍵策略包括:建立高品質的圖譜數據庫、優化數據採集參數以及數據分析策略都需要進一步開發及應用。在本研究中，我們的目標是通過以下方法實現深度人類蛋白體的定量：（i）在複雜樣品中混入標準樣品以及利用三物種的混合樣品中評估DIA的在定量上的表現;（ii）建構全面性的肺癌圖譜數據庫（iii）實際將single-shot DIA應用於肺癌組織進行蛋白體分析進行癌症生物學的研究。 在論文的第一部分，我們利用了將酵母菌當作背景的UPS1標準品和三物種混合而成蛋白質樣品，評估DIA跟DDA以及多反應監測（MRM）的分析性能。就定量準確性以及再現性而言，DIA的定量表現比DDA方法更傑出，同時在定量線性結果以及穩定性能夠達到跟MRM匹配的水準。接下來為了能夠建立足以進行深入的蛋白質體分析的肺癌圖譜數據庫，我們挑選了五個非小細胞肺癌細胞株和收集來自22個肺癌病人的肺腫瘤組織樣品，希望結合高鹼性逆相層析(high-pH reversed phase fractionation)降低樣品複雜度進而提高整體蛋白質涵蓋率，並且利用Orbitrap Fusion Lumos MS在DDA模式下進行採集了191筆數據後，我們成功的構建迄今為止規模最大的人類蛋白體圖譜數據庫其中含有12,377個蛋白質以及237,701個胜肽。並且透過這個圖譜資料庫的比對，我們可以利用single-shot的DIA方法，在HeLa細胞株中定量分析到7,143個蛋白。 在論文的第三部分，我們實際利用single-shot DIA搭配建立好的肺癌圖譜庫進行分析，針對來自16名患者的肺腫瘤和周圍正常組織進行定量分析以及比較。總共8,920個蛋白質組在1％蛋白及前驅物離子的偽陽性率(FDR)水平下成功被鑑定，平均每個組織可以被定性到7,468個蛋白質組（5,956個蛋白三重複定量低於20％CV）。來自32個組織樣品的定量表現證明了高品質的定量穩定性。腫瘤與鄰近正常組織之間進行成對樣本t檢定後，我們獲得了4,586個顯著差異表達的蛋白（FDR <0.05）。 2,954個正調控蛋白和1,632個負調控蛋白分別在DAVID數據庫上進行路徑富集分析。所有富集路徑（FDR <0.05）可分為四組與癌症標誌相關的類別：促進細胞生長的訊息持續活化，腫瘤的侵犯和轉移，癌細胞能夠永遠複製和重整能量代謝之方式。整體的結果與目前癌症研究報導有高度一致性。 總結來說，這項研究提供了迄今為止最深的蛋白質圖譜資料庫，並且能夠搭配single-shot DIA進行免標記定量。在本篇研究中說明了DIA展現的高靈敏度以及蛋白質體覆蓋率還有高準確的定量表現能夠針對非常有限的臨床樣品提升蛋白質體學的分析效能。

Mass spectrometry-based proteome analysis has become a powerful tool to advance our molecular understanding of complex biological processes in cell, tissues and other clinical samples. However, in-depth proteome quantitation usually requires multiple peptide fractionations, large amount of starting sample and extended data acquisition and processing, which limits application to clinical specimens with limited sample amount. Besides, the data-dependent acquisition (DDA) is commonly used in LC-MS/MS and its semi-stochastic nature of DDA results in poor run-to-run reproducibility in quantitative proteomics. The alternative approach, data independent acquisition (DIA), was recently introduced aiming parallel fragmentation of all ions and coupled to previously acquired and stored mass spectra reference library. It has been become a promising strategy combining advantages of discovery power of DDA with quantitative accuracy of targeted multiple reaction monitoring (MRM). However, coverage at the genome-depth still awaits to be improved. To improve the profiling coverage by DIA approach, we consider few critical strategies, including establishment of high quality reference library, optimum data acquisition parameter and the data analysis pipeline requires further development for application. In this study, we aimed to achieve deep human proteome quantitation by (i) evaluating quantitation performance of DIA using standard samples spiked to complex background as well as hybrid complex proteome; (ii) constructing a comprehensive lung cancer library and (iii) applying single-shot DIA to lung cancer tissues proteome analysis to study the cancer biology. In the first part of thesis, we evaluate the analytical performance of DIA and compare to the well-established DDA and multiple reaction monitoring (MRM) by standard UPS1 spiked to yeast sample and hybrid proteome from three species. The DIA has outperformed to the state of the art DDA approach in terms of quantitation accuracy and reproducibility while resulting in comparable quantitation linearity and robustness to targeted MRM method. To further improve the depth of proteome profiling, we nextconstructed a comprehensive library using five NSCLC cell lines and pooled lung cancer tumor tissues from twenty-two patients. By using high pH reverse phase fractionation followed by Orbitrap Fusion Lumos MS data acquisition, we were able to build by far the most comprehensive reference library with 12,377 protein groups (237,701 peptides) from 191 raw files. Using this library, single-shot DIA analysis is able to generate 7,143 proteins in HeLa cell line. In the third part of thesis, we applied single-shot DIA analysis for quantitative comparison of paired lung tumors and adjacent normal tissue from 16 patients. In total, 8,920 proteins were identified at 1% false discovery rate (FDR) at protein and precursor level Among them, average of 7,468 protein groups identified per tissue (5,956 quantified below 20% CV among triplicates). The quantitation performance from the 32 tissue samples demonstrated high-quality quantitation and robustness. After pared t-test between tumor and adjacent normal tissue, we got 4,586 significant differentially expressed proteins (FDR<0.05). 2,954 up-regulated and 1,632 down-regulated proteins followed by pathway enrichment analysis processing on DAVID database. All enriched pathways (FDR<0.05) can be categorized into 4 groups related to cancer hallmarks: sustaining proliferative signaling, activating invasion & metastasis, enabling replicative immortality and deregulating cellular energetics. The overall outcomes revealed high consistence with published cancer report. Overall, this study provides so far the deepest proteome mass spectra library resource for utility of single-shot DIA for label free quantitation．The demonstrated sensitivity, proteome coverage and high accuracy reveal its higher performance for clinical proteomics, especially for clinical specimens with limited sample amount.