LncRNA de novo discovery reveals noncoding RNAs as major molecular mechanism associating coronary artery disease GWAS variants with causal genes to confer disease risk

Long noncoding RNAs (lncRNA) comprise an underlying regulatory network in the human genome that remains largely unexplored and could represent a molecular mechanism connecting GWAS risk variants with the disease causal genes. Human coronary artery smooth muscle cells (HCASMC) are one of the most important cells in the regulation of atherosclerotic disease progression. In this study, we aimed to discover the HCASMC-specific lncRNA collection and to explore the mechanistic relationship between HCAMSC lncRNAs with increased disease risk for coronary artery disease. We applied 6 different stimuli relevant to athero-condition, including stimulations with TGFbeta, TNFalpha, PDGFD and serum, and two transcription factor knock-downs TCF21 and SMAD3, and performed deep RNA sequencing on 48 HCASMC samples. We designed a lncRNA discovery pipeline to maximize the detection of tissue specific and low expressed lncRNAs. This generated a set of 53076 lncRNAs, that showed increased association with CAD GWAS variants, various HCASMC eQTL data sets and GTEx eQTLs for tissues enriched in smooth muscle. Module analysis revealed lncRNAs highly associated with TGFbeta and general pro-differentiation traits located in the CAD GWAS loci, such as, FES/FURIN, COL4A1/A2, CDKN2A/B, TGFB1, and FN1. Transcription factor motif analysis revealed the presence of HCASMC relevant factors, such as, TCF21, ZEB1, ZEB2, JUN, JUND, and an SRF cofactor ELK4, near the boundaries of HCASMC lncRNA. We defined lncRNA QTLs using the GTEx Coronary Artery dataset, and showed colocalization with other HCASMC QTLs, such as expression QTLs, chromatin looping QTLs, chromatin accessibility QTLs and TCF21 binding QTLs. We show several examples of CAD GWAS loci where lncRNAs show regulatory function, such as FES/FURIN, FN1, COL4A1/A2 and TGFB1. We unraveled a complex network of regulatory interactions at FES/FURIN locus, involving TCF21 and lncRNA 43779.9, and present a model in which TCF21 inhibited lncRNA 43779.9 regulates FES gene and indirectly the pro-differentiation FURIN gene by creating a looping interaction with the intron 1 of FES transcript. We define 5 regulatory variants at the FES/FURIN locus and propose a causal variant, rs35346340, that disrupts TCF21 binding and subsequently influences 43779.9 expression. Finally, we show the presence of lncRNAs in deeply sequenced TCF21 pooled ChIPSeq and discover TCF21 binding lncRNAs, 3938.1 and 3852.1 as ACTA2-regulating transcripts. This study defines lncRNAs as essential regulators of GWAS loci and shows the importance of using deep sequencing approach to further explore the genomic regulatory landscape of lncRNAs in human tissues.