Running title: Barrett's oesophagus and OAC share a primitive intestinal transcription factor network.
AbstractOesophageal adenocarcinoma (OAC) is one of the most frequent causes of cancer deaths and yet compared to other common cancers, we know relatively little about the molecular composition of this tumour type. To further our understanding of this cancer we have used open chromatin profiling to decipher the transcriptional regulatory networks that are operational in OAC. We have uncovered a transcription factor network that is usually found in primitive intestinal cells during embryonic development, centred on HNF4A and GATA6. These transcription factors work together to control the OAC transcriptome. Importantly, we show that this network is activated in Barrett's oesophagus, the putative precursor state to OAC thereby providing novel molecular evidence in support of stepwise malignant transition. Furthermore, we show that HNF4A alone, is sufficient to drive chromatin opening and activation of a Barrett's-like chromatin signature when expressed in normal human epithelial cells. Collectively, these data provide a new way to categorise OAC at a genome scale and implicate HNF4A activation as a potential pivotal event in regulating its malignant transition from healthy cells.
Results
Identification of a network of transcription factors active in OACPreviously we used ATAC-seq to profile the open chromatin landscape of OAC cell lines and identified AP1 as an important transcription factor family in controlling the transcriptional networks in OAC . To further interrogate the transcriptional networks operating in OAC, we decided to take an alternative approach, starting with clearly defined open chromatin datasets from OAC patient samples rather than a diverse set of OAC-derived cell lines. We previously validated our cell line-derived results by profiling the open chromatin of six patient-derived biopsies but this revealed two distinct subclusters of OAC samples based on their open chromatin landscapes; one that clustered with normal samples and one that was unique . We revisited this issue by performing subclustering by PCA analysis following the addition of an additional paired normal and OAC dataset and again we observed a clear partitioning of samples ( Supplementary Fig. S1A). We therefore re-focussed our attention on the four OAC samples which are clearly distinct from the normal samples (T_002, T_003, T_005, T_006; Supplementary Fig. S1A).To derive a unified dataset of open chromatin regions in oesophageal-derived tissue, we combined all the ATAC-seq data from three normal oesophageal tissue samples (matched with T_003, T_005 and T_006) and these four OAC samples and recalled the open accessible chromatin regions from this combined data. We selected the top 50,000 open regions ( Supplementary Table S1A) and found that these were roughly equally divided between promoter-proximal, intragenic and intergenic regions (Fig. 1A). PCA analysis based on these peaks confirmed the distinct clustering of the normal an...