Jankowski Aleksander, Obara Paulina, Mathur Utsav, Tiuryn Jerzy
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw, Banacha 2, Warsaw, 02-097, Poland.
Current address: Genome Biology Unit, European Molecular Biology Laboratory, Meyerhofstraße 1, Heidelberg, 69117, Germany.
BMC Syst Biol. 2016 Feb 4;10:14. doi: 10.1186/s12918-016-0258-3.
The enhanceosome is an enhancer located upstream of the human interferon β gene, bound by transcription factor (TF) complex of extremely rigid structure. Within these rigid constraints, even a slight change of distances between transcription factor binding sites (TFBS) results in loss of functionality of the enhanceosome. We hypothesized that smaller subunits of the enhanceosome may entail TF complex formation in other regulatory regions.
In order to verify this hypothesis we systematically searched for dimerization preferences of the TFs that have TFBS in the enhanceosome. For this we utilized our recently developed tool, TACO. We performed this computational experiment in a cell-type-specific manner by utilizing cell-type-specific DNase-seq data for 105 human cell types. We also used 20 TRANSFAC motifs comprising not only the usual TFs constituting the enhanceosome but also the architectural proteins of High Mobility Group I(Y) (HMG I). A similar experiment used 42 DNase-seq data sets for mouse cell types. We found 137 statistically significant dimer predictions in the human genome, and 37 predictions in the mouse genome, that matched the positioning on the enhanceosome with ±2 bp tolerance. To characterize these predicted TF dimers, we performed functional analysis (Gene Ontology enrichment) for sets of genes which were in the neighbourhood of predicted dimer instances. A notable feature of these instances is that (1) most of them are located in introns of genes, (2) they are enriched in regulatory states, and (3) those instances that are located near transcription start sites are enriched for inclusion in computationally predicted enhancers. We also investigated similarity of dimer predictions between human and mouse.
It follows from our experiments that, except for homodimer formed by IRF proteins, the rest of the dimers were formed exclusively between one of the transcriptional activators (ATF-2/c-Jun and IRF) and a HMG I protein. NF- κB did not participate in forming dimers with other proteins. Dimers predicted in mouse were fully contained in those predicted in human, with exactly the same spacing and orientation. Intriguingly, in most of the cases the enhanceosome motifs have 1 bp wider spacing than the corresponding dimers predicted genome-wide, which is likely caused by the overall 3D structure constraints of the enhanceosome-bound complex.
增强体是位于人类干扰素β基因上游的一种增强子，由结构极其刚性的转录因子（TF）复合物结合。在这些严格的限制条件下，转录因子结合位点（TFBS）之间哪怕距离稍有变化都会导致增强体功能丧失。我们推测增强体的较小亚基可能在其他调控区域参与TF复合物的形成。
为了验证这一假设，我们系统地搜索了在增强体中具有TFBS的TF的二聚化偏好。为此，我们利用了我们最近开发的工具TACO。我们通过利用105种人类细胞类型的细胞类型特异性DNase-seq数据，以细胞类型特异性的方式进行了这个计算实验。我们还使用了20个TRANSFAC基序，其中不仅包括构成增强体的常见TF，还包括高迁移率族蛋白I（Y）（HMG I）的结构蛋白。类似的实验使用了42个小鼠细胞类型的DNase-seq数据集。我们在人类基因组中发现了137个具有统计学意义的二聚体预测结果，在小鼠基因组中发现了37个预测结果，这些结果与增强体上的定位匹配，容差为±2 bp。为了表征这些预测的TF二聚体，我们对预测二聚体实例附近的基因集进行了功能分析（基因本体富集）。这些实例的一个显著特征是：（1）它们大多数位于基因的内含子中；（2）它们在调控状态中富集；（3）那些位于转录起始位点附近的实例在计算预测的增强子中富集。我们还研究了人类和小鼠之间二聚体预测的相似性。
从我们的实验可以得出，除了由IRF蛋白形成的同二聚体外，其余的二聚体仅在一种转录激活因子（ATF-2/c-Jun和IRF）与一种HMG I蛋白之间形成。NF-κB不参与与其他蛋白形成二聚体。在小鼠中预测的二聚体完全包含在人类中预测的二聚体中，具有完全相同的间距和方向。有趣的是，在大多数情况下，增强体基序的间距比全基因组预测的相应二聚体宽1 bp，这可能是由与增强体结合的复合物的整体三维结构限制导致的。
