High-Content Image Analysis of Global DNA Methylation Patterns for Drug Screening

描述（由申请人提供）：癌症是哺乳动物细胞中遗传变化（基因突变，染色体失衡）和表观遗传变化（DNA甲基化和组蛋白甲基化/乙酰化）积累的结果。与遗传变化相比，表观遗传变化是可逆的，因此在新的表观遗传癌症治疗药物发现中提供了一个有希望的靶点。去甲基化药物引起整体DNA甲基化的变化，这可以通过光学显微镜观察到细胞核内基因组的大规模重组。这些变化可以通过高分辨率（HR）三维（3-D）荧光显微镜测量单个细胞甲基化DNA的显著差异分布模式。因此，定量DNA甲基化成像（QDMI）将成为表观遗传治疗中表征癌细胞的重要工具，用于药物发现和个性化癌症诊断和预后。据我们所知，目前还没有高含量（HC）信息系统支持基于HR - 3-D成像的癌细胞评估，并自动分析差异DNA甲基化。为了推进这一领域超越现有的HC解决方案，我们提出了新的基于统计的工具，将其集成到甲基化模式识别工作流程中。设想的系统将执行以下任务：(i)核DNA甲基化模式和特征的自动提取，（ii）通过Kullback-Leibler的分歧来评估细胞培养的差异性，以及（iii）基于特征的分类来区分高甲基化DNA和低甲基化DNA的细胞核。细胞培养中的药物反应将作为最后两个步骤的综合产物进行评估，并以叠加在原始图像上的伪彩色图进行可视化，以供专家进行视觉和定量评估。系统性能将在前列腺癌细胞系DU145上进行测试和验证，使用三种抗癌药物：阿霉素、5-阿扎胞苷、丙戊酸，以不同剂量治疗7天。一个专门的数据库将用于存档、链接和整合图像、药物、治疗和定量数据。HC信息系统将支持基于图像的DNA甲基化分析从成熟的技术发展为诊断系统。我们的目标是为快速发展的研究领域带来根本性的贡献，该领域结合了生产快速、自动化的基于图像的技术所需的多学科努力。目标是在天然环境中对数千种哺乳动物细胞进行高通量表观遗传筛选，其长期潜在益处包括：（一）寻找表观遗传治疗的靶点并预测对癌症治疗的反应性；（二）评估影响细胞表观基因组组成的营养和环境因素；（三）在细胞基础上表征复杂的表观基因组相关疾病；（四）使细胞模型能够用于表观遗传药物开发。

DESCRIPTION (provided by applicant): Cancer is the result of accumulations of genetic changes (gene mutations, chromosomal imbalances) and epigenetic changes (DNA methylation and histone methylation/acetylation) in mammalian cells. In contrast to genetic changes, the epigenetic changes are reversible, thus presenting a promising target in drug discovery for novel epigenetic cancer therapies. Demethylating drugs cause changes in global DNA methylation that can be visualized as large reorganizations of the genome within a cell's nucleus by light microscopy. These changes can be measured as significant differential distribution patterns in the methylated DNA of individual cells using high-resolution (HR) three-dimensional (3-D) fluorescence microscopy. Thus, quantitative DNA methylation imaging (QDMI) will be a vital tool in the characterization of cancer cells in epigenetic therapy, for both drug discovery and personalized cancer diagnostics and prognostics. To the best of our knowledge there is no high-content (HC) informatics system supporting HR 3-D imaging-based evaluation of cancer cells, with automatic analysis of differential DNA methylation. For advancing this field beyond existing HC solutions, we propose our new statistics-based tools to be integrated into the methylation pattern recognition workflow. The system envisaged will perform the following tasks: (i) automatic extraction of nuclear DNA methylation patterns and features, (ii) evaluation of cell culture dissimilarity by means of Kullback-Leibler's divergence, and (iii) feature based classification to distinguish nuclei with hypermethylated DNA and hypomethylated DNA. The drug response in the cell culture will be assessed as a combined product from the two last steps, visualized as pseudo-colored maps superimposed on the original images for experts' visual and quantitative assessment. System performance will be tested and validated on the prostate cancer cell line DU145, treated with three anticancer drugs: doxorubicin, 5-azacytidine, and valproic acid, with different doses for seven days. A dedicated database will be used to archive, link and integrate image, drug, treatment and quantitative data. The HC Informatics System will support the evolution of image-based DNA methylation analysis from a proven technique into a diagnostic system. We aim to bring a fundamental contribution to a quickly growing research field that combines multidisciplinary efforts required for producing fast, automated image-based techniques. The goal is high-throughput epigenetic screening of thousands of mammalian cells in their native environment, with long-term potential benefits including: (i) finding targets for epigenetic treatment and predicting responsiveness to cancer therapy, (ii) assessing nutritional and environmental factors that impact the epigenomic makeup of cells, (iii) characterization of complex epigenomics-related diseases on a cellular basis, and (iv) enabling the use of cellular models in epigenetic drug development. PUBLIC HEALTH RELEVANCE: The aim of this research is to develop a high-content analysis method for imaging-based cancer cell screening and evaluation of drug effects in epigenetic cancer treatment. Automatic analysis and quantification of global DNA methylation patterns in individual cells allows the evaluation of drug response and treatment effects.