Hotspots of aberrant enhancer activity in fibrolamellar carcinoma reveal molecular mechanisms of oncogenesis and intrinsic drug resistance

Timothy A. Dinh; Ramja Sritharan; F. Donelson Smith; Adam B. Francisco; Rosanna K. Ma; Rodica P. Bunaciu; Matt Kanke; Charles G. Danko; Andrew P. Massa; John D. Scott; Praveen Sethupathy

Our gene window analyses (Fig. 3) identified genes that may be regulated by FLC-specific enhancers. However, nearby genes may have the same TREs within their windows. In order to more confidently link individual FLC-specific enhancers with putative gene targets, we correlated enhancer activity to gene transcription levels across all FLC tumors, as described previously (Corces et al., 2018) . ChRO-seq allows us to quantify both enhancer activity and gene transcription from a single experimental dataset, thereby avoiding confounding variables that can arise when using multiple different assays. Correlations between transcriptional activity of enhancers and genes within 100 kb of each other were compared to a null distribution of inter-chromosomal enhancer-gene pairs to calculate p-values (Fig. 4A ,B). Windows larger and smaller than 100 kb had reduced power to detect statistically significant gene-enhancer correlations (Fig. S3A ). Using an FDR < 0.1, we linked 1697 FLC-specific enhancers to putative target genes ( Fig. 4B ,C). As expected, we observed that the frequency of predicted geneenhancer links decreases with increasing distance between them (Fig. 4D ). We found that most enhancers are linked to only one or two genes (mean = 1.33, Fig. 4E ) and most genes are linked to only one or two enhancers (mean = 1.78, Fig. 4F ). The top 5% of genes (in terms of enhancer connectivity) are each linked to at least 4 enhancers (Fig. 4G , Table S5 ). The top 5% includes FAM19A5, LINC00473, VCAN, SLC16A14 (Fig. 4H ) and CA12 (Fig. 4I) , which are putatively linked to 20, 9, 9, 9, Fig. S4 ; Table S6 ). Interestingly, while CA12 is not highly transcribed in NML, there is a substantial amount of ChRO-seq signal at the promoter (Fig. 4I) , indicative of polymerase pausing. This is unlike what we observe at the SLC16A14 locus (Fig. 4H) , where there is no ChRO-seq signal in NML even at the TSS. This observation suggests that transcriptional pausing may be another mechanism that regulates CA12 expression. Therefore, CA12 may be poised for expression, whereas the SLC16A14 locus is completely inactive in NML and dramatically rewired for activation in FLC.