TGFβ-SMAD信号通路是掌控胚胎干细胞命运的核心信号之一，但是其调控干细胞分化的机制尚不清楚。我们通过全基因组的ChIP-seq 和RNA-seq分析,发现Smad2/3, Smad4和Trim33共同占据并直接转录调控绝大部分中内胚层基因的表达。我们发现在启动子和增强子上均有经典的Smad3/4结合位点，并在启动子上找到了潜在的Smad2结合序列。这一发现与目前认为Smad2蛋白由于3号外显子（Exon 3，E3）的插入而无法与DNA结合的观点相悖。通过CRISPR技术，我们证实Smad2的E3是决定鼠胚胎干细胞中内胚层分化的关键。基于以上重要发现，我们提出Smad2-Smad3-Smad4三聚体可以环状连接分化基因的启动子和增强子，从而促进中内胚层调控基因转录和干细胞分化。因此，本课题将分析Smad复合体与分化基因的启动子和增强子结合的结构基础，以及调控干细胞分化的分子机制。

The TGFβ family of cytokines and their SMAD signal transduction pathway are central to metazoan development and tissue regeneration. Disruption of TGFβ signaling underlies many diseases, both inherited and somatic. The basic mechanism of TGFβ-SMAD signal transduction from the membrane to the nucleus is understood to a degree, but the cardinal role of TGFβ-SMAD in embryonic stem cell (ESC) differentiation remains poorly understood. .. In order to acquire the necessary data and conceptual framework for the present work, we launched a genome-wide ChIP-seq and RNA-seq analysis of Smad and Trim33 proteins in mouse ESCs. We defined that the differentiation gene responses require nodal-driven mediators Smad2/3, Trim33 and Smad4. According to their genome occupancy maps in ESCs and d2.5 embryonic bodies, interestingly, the highest density of Smad and Trim33 sequence tags in most target genes occurs not at proximal promoters but at distal sites located up to 50kb from the transcriptional start site (TSS). These sites near mesendoderm specification genes are frequently co-occupied by Smad2/3, Smad4 and Trim33. Taking Goosecoid (Gsc) as a template, a model Smad-regulated master differentiation gene， We found proximal GC-rich motif and the distal CAGAC motifs are essential for Gsc transcription and act as potential binding sites for Smad recognition. These findings laid basic foundation to the question of how the SMAD transcriptional complexes target master differentiation genes for regulation. We hypothesized that the Smad2-Smad3-Smad4 trimer would link the promoter and distal enhancer region of mesendoderm differentiation genes. We will address the structural determinants of Smad recognition of these master differentiation genes, test a model for the trimeric nature of Smad complexes and probe the model functionally in ESCs and in mice. We additionally challenge the long-held dogma that Smad2 ¬–the most essential of all Smad family members in vertebrates– does not bind DNA. This knowledge will help explain many extant puzzles in the TGFβ field and beyond.