A Boolean approach for novel hypoxia-related gene discovery

Abstract

Hypoxia plays a major role in the etiology and pathogenesis of most of the leading causes of morbidity and mortality, whether cardiovascular diseases, cancer, respiratory diseases or stroke. Despite active research on hypoxia-signaling pathways, the understanding of regulatory mechanisms, especially in specific tissues, still remain elusive. With the accessibility of thousands of potentially diverse genomic datasets, computational methods are utilized to generate new hypotheses. Here we utilized Boolean implication relationship, a powerful method to probe symmetrically and asymmetrically related genes, to identify novel hypoxia related genes. We used a well-known hypoxia-responsive gene, VEGFA, with very large human expression datasets (n = 25,955) to identify novel hypoxia-responsive candidate gene/s. Further, we utilized in-vitro analysis using human endothelial cells exposed to 1% O2 environment for 2, 8, 24 and 48 hours to validate top candidate genes. Out of the top candidate genes (n = 19), 84% genes were previously reported as hypoxia related, validating our results. However, we identified FAM114A1 as a novel candidate gene significantly upregulated in the endothelial cells at 8, 24 and 48 hours of 1% O2 environment. Additional evidence, particularly the localization of intronic miRNA and numerous HREs further support and strengthen our finding. Current results on FAM114A1 provide an example demonstrating the utility of powerful computational methods, like Boolean implications, in playing a major role in hypothesis building and discovery.

Fig 1. Boolean implication analysis to identify novel hypoxia-responsive genes.

(a) Schematics of approach utilized to filter hypoxia-related genes. We used VEGFA as the seed gene to identify genes responding to hypoxia. Filter includes ranking, conserved relationship in mice and novelty. (b—e) Scatter plots depicting the expression profile of 25,955 samples for VEGFA (probset 210512_s_at, x-axis) and the additional VEGFA probset 210513_s_at (c) and top correlated genes i.e., SDC4 (d), ERRFI1 (e) and ITGAV (f). (f—i) Validation of candidate seed gene (VEGFA) and SDC4 (h), ERRFl1 (i) and ITGAV (j) in HPAECs in normoxia (21% O₂) and at 2, 8, 24 and 48 hours of constant 1% O₂. (*, P<0.05).

Fig 2. Identification of FAM114A1 as a novel hypoxia responsive gene.

(a) Scatter plot depicting the expression profile of VEGFA (probe ID, 210512_s_at; x-axis) and FAM114A1 (probe ID, 226697_at; y-axis) in n = 25,955 ‘human’ samples. (b) Scatter plot depicting the expression profile of Vegfa (probe ID, 1420909_at; x-axis) and Fam114a1 (9130005-N14Rik, probe ID, 1417272_at; y-axis) in n = 11,758 ‘mice’ samples (Correlation = 0.451606). (c) RT-PCR results depicting expression level of FAM114A1 under normoxia (21% O₂) and at 2, 8, 24 and 48 hours of constant 1% O₂ measured in HPAEC. Insert, gene expression dynamics of VEGFA and FAM114A1 along the boolean path. (*, P<0.05).

Fig 3. Distinctive properties of FAM114A1 and its role in hypoxia.

(a) FAM114A1 is expressed in most of the tissues but its expression is low in the nervous system. (b) Position of hypoxia-response elements (HREs) consenses in FAM114A1 and intronic miRNA, miR-574 in intron-1. The region constituting miR-574 is evolutionarily conserved. (c) Network analysis reveals physical interaction of FAM114A1 with AKT1S1 (AKT1 substrate 1) and N4BP1 (NEDD4 binding protein 1).