結合顯微鏡的單細胞多體學分析技術解析癌症子族群

Abstract

癌症細胞並不是一整塊都由相同的細胞所組成，反之它們有著各種相差甚遠的表型和功能 (統稱為異質性 heterogeneity)，例如不同的腫瘤侵略程度或受治療的敏感性，這最終將會影響治療是否能成功和預後結果。然而在理解到每顆腫瘤細胞都有所不同之前，我們解讀生物變化量的意涵可能存在偏誤，包括 DNA 序列 (基因體學)、基因活動 (轉錄體學)、 DNA 化學修飾 (表觀遺傳學)、蛋白質含量或轉譯後修飾 (蛋白質體學) 和代謝物 (代謝體學)。為了剖析異質的所有細胞，目前已經開發了數種單細胞測序技術及所屬的單細胞體學 (single-cell omics)，除了全面性地顯示出腫瘤內部有多異質，更揭示出癌細胞子族群 (subpopulations) 中獨特的分子特徵。然而，現有的單細胞測序與癌細胞實際的功能行為或表型 (phenotypes) 沒有直接關係，因此透過單細胞分析得到的子族群狀態或分子特徵無法輕易地與其表型異質性進行連結。在本論文中，我們旨在開發一個完整的分析方法來填補這個不足，並透過使用各種單細胞分析技術對不同表型進行分析來證明其成效，包括惡性遷移細胞的單細胞轉錄體學分析 (第二章)、異常 DNA 損傷反應細胞的單細胞蛋白質體學分析 (第三、四章) 以及多重細胞分裂的單細胞基因體和轉錄體分析 (第五章)。在這些章節中我們使用功能性單細胞分析技術 (Functional single-cell profiling) 提供了各種異質癌症相關問題的可能解答。 我們首先在第一章統整了研究腫瘤異質性的重要性，並探討當前可用於分析腫瘤異質性的技術。接著，我們深入討論各種有研究子族群潛力的方法，檢視其能否找到與細胞功能相關的體學數據，候選的方法包括原位成像並測序和單細胞分離結合體學分析技術。此外，我們還探討了各種單細胞分析技術的原理，並評估了它們在解決這個具有挑戰性的問題方面的優缺點。 為了找到表型和基因型之間缺少的拼圖，我們在第二章介紹了我們的功能性單細胞選擇 (functional single-cell selection, fSCS) 技術。它結合了使用超寬視野光學顯微鏡 (ultra-wide field-of-view optical microscope, UFO microscope) 的活體細胞顯微鏡錄像、即時圖像分析和靶向單個細胞的光選擇。我們進一步將其與最先進的單細胞 RNA 測序技術結合起來，統稱為功能性單細胞 RNA 測序 (FUNseq)。透過這項技術，我們成功地分析了混合人類乳腺 MCF10A 細胞中最惡性遷移的細胞子族群，並發現上皮細胞間質轉化相關基因的活化。這類基因已知在細胞遷移扮演重要功能，因此也進一步證實了顯微鏡可觀察的動態表型和潛在差異表達的轉錄體之間強烈的相關性。 在 FUNseq 成功連結動態表型與轉錄體之後，我們進一步將 FUNseq 分析技術從單細胞 RNA 測序擴展到單細胞蛋白質體學，因為蛋白質體學可以提供直接生物功能的蛋白質活性訊息。我們在第三章描述了這項功能性單細胞蛋白質體學 (Functional single-cell proteomics, FUNpro) 分析技術，我們解決了先前利用質譜分析單細胞蛋白質體學時缺少的空間及時間資訊，完美地填補了我們理解蛋白質體異質性方面的空白。本章闡述了用於進行 FUNpro 分析時完整且詳細步驟說明，包含 fSCS 及質譜單細胞蛋白質體學 (SCoPE-MS) 分析。除此之外，我們也提供常見除錯步驟於此 FUNpro 技術指南，希望能讓更多研究人員能夠使用該技術來深入研究腫瘤細胞的動態行為來更廣泛地了解蛋白質體異質性。 癌症復發通常是由腫瘤子族群內能抵抗癌症治療的細胞催生，而這些細胞主要表現出異質性 DNA 損傷反應 (DNA damage response)，因此類癌細胞可能得以更有效地修復 DNA 損傷，並抵消癌症治療 (例如放射和化學治療) 的療效。為了深入了解箇中奧秘，我們在第四章中利用 FUNpro 技術研究電離輻射後 DNA 損傷反應的異質性。利用 FUNpro 分析技術，我們同步觀測超過 3,000 個來自人類 U2OS 細胞的 DNA 損傷反應，並同時使用特別開發的細胞圖像分析演算法即時追蹤每顆細胞的 DNA 損傷相關蛋白質聚集的形成與消散。令人驚訝的是，並非所有的細胞在照射輻射後啟動相同或既定的 DNA 損傷反應途徑以恢復染色體完整性，實際上十分之一細胞中的 DNA 損傷蛋白質聚集持續無法消散，這推翻了原先預期的上升之後消散的波動趨勢。我們接著對這些特異細胞進行光選擇、分離和單細胞蛋白質體分析後，顯示了染色體凝聚蛋白 PDS5A 在減輕延長 DNA 損傷方面的重要作用。這一發現推進了我們對複雜的 DNA 損傷反應機制的理解，並強調了探索 DNA 損傷反應異質性對於制定有效治療策略對抗具有抗藥性的癌細胞的重要性。 除了 DNA 損傷反應之外，基因重組和染色體異常所導致的染色體不穩定性 (chromosomal instability, CIN) 也是腫瘤演化和抗治療性的一個重要驅動因素。然而，由於極其罕見，自然發生的不穩定染色體事件仍然知之甚少，限制了對其機制的深入了解。為了解決這個問題，我們在第五章介紹了 CIN-seq，可以對表現出不穩定染色體表型的稀有細胞子族群進行特異性分析，特別是罕見的多重細胞分裂 (multipolar mitosis)。在這類罕見的事件中，一個細胞分裂成至少三個子細胞，而不是典型的兩個。應用 CIN-seq，我們分別分析了這些三分裂細胞的單細胞基因體學和轉錄體學。我們的結果揭露了許多細胞內的染色體複製數變異 (copy number variation) 呈現混亂並破碎，證實了多重細胞分裂過程會隨機分離他們的染色體。此外，單細胞 RNA 測序結果鑑定出非活化的磷酸酶和張力素同源物 PTEN 是三分裂細胞的獨特分子特徵，在抑制劑試驗下進一步驗證了其確可驅使多重細胞分裂。此外，本章探討了此神秘三分裂背後的新分子機制，包括先天免疫反應、細胞質分裂失敗和 16 號染色體異常增加，這些發現為尋找新的治療策略奠定了可貴的基礎。 第六章總結了我們利用 fSCS 分析技術和相對應的單細胞測序方法 (即 FUNseq、FUNpro 和 CIN-seq) 研究惡性遷移、異常 DNA 損傷反應和多重細胞分裂，其結果在表型異質性和相對應分子機制上讓我們獲得許多寶貴的見解。此外，我們探討了可以最佳化分析技術的步驟和並提供可參考的技術改進，包括活體細胞成像、訊噪比和各種圖像分析演算法技術。本章還提供了一些可行的替代技術，例如使用多重光選擇技術以提高 fSCS 分析技術的準確度和通用性。此外，我們概述了得以推進和改進這些技術的未來方向，希望能持續對腫瘤異質性有更完整且深入的理解。 本論文完整地描述了功能性單細胞選擇分析技術 (fSCS) 及其在研究各種不尋常生物事件中的應用，並透過這些研究成果，我們對在顯微鏡下的動態表型與其相對應的分子訊息更加了解。同時圖像分析演算法、光學硬體和樣品製備流程也有許多重大改進，進一步使分析技術更加穩定可靠。我們希望未來能持續使用這個強大的分析技術探索各種細胞表型，並闡明各種腫瘤異質性相關的問題。最終，我們希望這些進步將讓我們更深入了解癌症機轉，並有朝一日能終結癌症治療失敗。

Tumor cells are not uniform masses of identical cells. In contrast, they exhibit various phenotypic and functional heterogeneities, including the aggressiveness of tumor progression or the susceptibility of the treatment, ultimately impacting therapeutic resistance and prognostic outcomes. This vast heterogeneity hinders our ability to correctly characterize variations in DNA sequences (genomics), gene activities (transcriptomics), chemical modifications on DNA (epigenetics), protein abundances or post-translational modifications (proteomics), and metabolites (metabolomics). To dissect individual cells amidst this complexity, several single-cell sequencing technologies have developed, illuminating comprehensiveness of heterogeneity within tumors and revealing unique molecular profiles across subsets of cancer cells. However, single-cell sequencing is independent to the functional behaviors (or phenotypes) of cancer cells, the molecular states captured through single-cell analysis lose direct connections with their phenotypic heterogeneities. In this thesis, we aimed to develop a comprehensive pipeline to address this deficiency and demonstrate its efficacy by profiling different phenotypes using various single-cell profiling techniques, including single-cell transcriptomic analysis of aggressively migrating cells (Chapter 2), single-cell proteomic analysis of cells with abnormal DNA damage responses (Chapter 3 and 4) and single-cell genomic and transcriptomic analyses of multipolar mitosis (Chapter 5). In these chapters, we showcased various biological questions we sought to answer with our comprehensive functional single-cell pipeline. We first summarized the importance of studying tumor heterogeneity and discussed currently available techniques to analyze tumor heterogeneity in Chapter 1. Furthermore, we delved into exploring the potential for studying phenotypic subpopulations by examining various methods that establish connections between cell functions and their underlying omics. These include in situ imaging/sequencing and single-cell isolation coupled with omic analysis pipelines. In addition, we explored the principles of various single-cell profiling techniques and evaluated their pros and cons in tackling this challenging issue. With the aim of bridging the gap between phenotypes and genotypes, we introduced our functional single-cell selection (fSCS) pipeline in Chapter 2. This integrates live-cell microscopy imaging using an ultrawide field-of-view optical (UFO) microscope with real-time image analysis and photoselection of targeted single cells. We further combined this with state-of-the-art single-cell RNA sequencing (scRNA-seq) technology, collectively termed as functional single-cell RNA sequencing (FUNseq). Through this technology, we successfully profiled the most aggressively migrating subpopulations of cells in intermingled human breast MCF10A cells, revealing upregulated epithelial-to-mesenchymal transition-related genes. This finding highlighted a strong correlation between microscopically-observable dynamic phenotypes and the underlying differentially expressed transcriptomes. Building upon the success of FUNseq in elucidating transcriptomes, we further expanded our FUNseq pipeline from scRNA-seq to single-cell proteomics (SCP), as proteomics can provide direct biofunctional indication of protein activity. We described this technology, functional single-cell proteomics (FUNpro), in Chapter 3, where we addressed the missing spatial/temporal information in previous mass spectrometry (MS)-based SCP analysis, bridging the crucial gap in our understanding of proteome heterogeneity. This chapter provided a comprehensive and detailed protocol for performing the fSCS and its integration with Single-Cell ProtEomics by Mass Spectrometry (SCoPE-MS). We presented a protocol consisting of live-cell imaging, real-time image analysis, targeted cells photoselection and isolation, as well as plate-based SCoPE-MS technology for proteomics data analysis. We offered readers a holistic guide for implementing the FUNpro technology, enabling researchers to gain a broader understanding of proteome heterogeneity through insights into the dynamic behaviors of tumor cells. Tumor relapse is often driven by treatment-resistant cells within a tumor subset displaying heterogeneous DNA damage responses (DDRs). Such cancer cells may repair DNA damage more efficiently and counteract the efficacy of cancer treatments such as radiotherapy and chemotherapy. To address the challenge, we aimed to study the DDR heterogeneity post ionizing radiation (IR) using the FUNpro technology in Chapter 4. While it was initially expected that all cells would initiate similar DDR processes upon irradiation, preserving chromosomal integrity through well-estabalished pathways. However, our results revealed a different story. Leveraging the FUNpro pipeline, we simultaneously screened > 3,000 DDR responses of human U2OS cells, and subsequently identified the formation/resolution of DDR protein foci for individual cells in real time using the dedicated intracellular image analysis algorithms. Surprisingly, we found that 10% of cells exhibited a prolonged and unresolved DDR foci trend, contradicting the expected up-&-down foci trend. Subsequent photoselection, isolation and single-cell proteomic analysis of these target cells unveiled the crucial role of the chromatid cohesion protein PDS5A in mitigating prolonged DNA damage. This finding advances our understanding of intricate DDR mechanisms and emphasizes the importance of exploring the DDR heterogeneity for effective treatment strategies against resilient cancer cells. In addition, chromosomal instability (CIN) is also a critical driver of tumor evolution and therapy resistance, driven by genomic reorganization and chromosomal abnormalities. However, spontaneously occurring CIN remains poorly understood due to its extreme rarity, limiting insights into its mechanisms. To address this, we introduced CIN-seq in Chapter 5, enabling specific profiling of rare cell subpopulations that exhibit live CIN phenotypes, particularly the unusual tripolar mitosis. In this rare event, a single cell divides into three daughter cells instead of the typical two. Applying CIN-seq, we respectively analyzed single-cell genomics and transcriptomics of these tripolar dividing cells. Our results revealed a chaotic pattern of copy number variation (CNV) across all chromosomes, confirming the random segregation during tripolar mitosis. Furthermore, scRNA-seq data identified the downregulation of the phosphatase and tensin homolog (PTEN) as the unique molecular signature of tripolar dividing cells, and its role in driving multipolar mitosis was further validated by inhibition assays. This chapter unravels the novel molecular mechanisms behind the mysterious tripolar mitosis, including innate immune responses, cytokinesis failure and chromosome 16 amplification. These findings pave the way of the identification of new therapeutic strategies. Chapter 6 consolidates our findings from employing the fSCS pipeline and the corresponding single-cell sequencing methods (i.e., FUNseq, FUNpro, and CIN-seq) to explore aggressive migration, abnormal DNA damage response, and tripolar mitosis. Our analyses unlocked insights into the heterogeneity of biological phenotypes and the underlying molecular mechanisms. Furthermore, we discussed perspectives for refining the pipeline and improving potential technical improvements, including live-cell imaging, signal-to-noise ratio, and diverse image analysis techniques. This chapter also proposed optimizations and the exploration of alternative technologies, such as the multiplex photoselection technique, to improve the precision and versatility of the fSCS pipeline. Additionally, we outlined future directions aimed at advancing the development and refinement of this pipeline, with the overarching goal of deepening our understanding of tumor heterogeneity. This thesis provides a comprehensive description of the functional single-cell selection pipeline, or fSCS, and its applications in studying rare and diverse biological events. Through these studies, we have gained valuable insights into the underlying molecular information and their corresponding phenotypes under the microscope. Numerous significant improvements in image analysis algorithms, optical hardware, and the sample preparation have been made to enhance the robustness of the pipeline. Our overarching goal is to utilize this comprehensive pipeline to explore a more in-depth analysis of various cellular phenotypes, shedding light on intriguing biological questions related to tumor heterogeneity. Ultimately, we hope that these advancements will contribute to a more profound understanding of cancer behavior and pave the way to overcome the challenges of cancer treatment failure.