Method Development and Applications of Low-input Proteomics

Proteomics offers complementary and more direct information to genomics and transcriptomics, essential for the understanding of complex biological processes. Low-input proteomics is performed when the available sample amount is limited, and it requires improvements in the whole workflow, including sample preparation, peptide separation, mass spectrometry (MS) detection, and data analysis, for higher sensitivity and accurate statistical analysis. In this thesis, several advances in low-input proteomics and its applications are described. First, an easy-to-use and scalable device for sample preparation, a 3-frit mixed-mode rare cell proteomic reactor (RCPR) for integrated processing low-input samples with minimized sample loss was developed. Using the 3-frit mixed-mode RCPR, 2 998±106 and 2 595±230 protein groups from 100 human embryonic kidney cells and 500 mouse cochlear hair cells, respectively, were identified, representing the best results so far in the literature for such low-input samples. Second, a fast and robust column fabrication method for narrow-bore capillary columns with negligible dead volume was developed, allowing the identification of an average of 3 043±39 protein groups from 1 ng of protein digest from human cells. Third, the sensitivity of the Q Exactive HF-X and timsTOF Pro mass spectrometers for nanogram-level samples was systematically optimized and improved by optimizing their data-dependent acquisition parameters. In a proof-of-concept application, we developed a two-step machine learning-based T cell subtyping strategy, and successfully extracted unique proteome classifiers for eight T cell subtypes for low-input samples collected from single multiple myeloma patients, despite different input cell numbers and individual differences. Collectively, a series of improvements in the whole workflow of low-input proteomics were made and their applications were demonstrated in two proof-of-concept applications.

与基因组学和转录组学相比，蛋白质组学提供互补的，更直接的信息，且蛋白质组学信息对理解复杂的生物学过程是必不可少的。当样品量很有限时，我们需要微量蛋白质组学，而且它需要在整个流程，从样品前处理，多肽分离，质谱检测，到数据分析都做出改进。本论文取得了一些关于微量蛋白质组学的进展和应用。首先，我开发了一个易用的，可塑的装置整合样品前处理装置-3个筛板的稀有细胞蛋白组反应器，能进行集成化的样品前处理，并使样品缺失降低到最小。使用该反应器，能分别从100个人胚肾细胞和500个小鼠耳蜗毛细胞中检测到2 998±106 和2 595±230个蛋白组，优于文献报道的结果。第二，我开发了一种方便可靠的窄内径毛细管制备方法，能从1纳克多肽样品中检测到3 043±39个蛋白组。第三，两种高端质谱，Q Exactive HF-X和 timsTOF Pro的质谱参数经过优化后，灵敏度得到了很大提升。最后，在一个应用中，我开发了一种两步法基于机器学习的T细胞分类策略，基于单个多发性骨髓瘤患者中的少量T细胞中检测到的蛋白信息，成功地提取出特征蛋白分类器用于区分八种主要T细胞。总结起来，本论文在整个微量蛋白质组学流程的每个主要步骤都做出了改进，并在两个应用中进行了展示。