In silico and in vitro tools to investigate molecular actors in lymphangiogenesis

Abstract

(Lymph)-angiogenesis, the creation of new lymphatic/blood vessels, plays an important role in cancer development and metastatic spread. While angiogenesis has been studied extensively over the years, particularly for its role in cancer, tumour lymphangiogenesis has only recently attracted interest. Tumour lymphangiogenesis leads to the activation of numerous signalling pathways. However, only a small proportion of the actor interactions and regulations at the origin of lymphangiogenesis occurring in this particular disease is currently understood. The development and use of in silico tools such as biological network inference offers a relevant approach to address these shortcomings. This work aims in investigating the transcriptomic molecular interactions behind lymphangiogenesis in a cancer context through gene regulatory network (GRN) inference and in vitro validation approaches. The GRN was constructed with machine learning algorithms using as input multi-perturbed gene expression data from publicly available studies in repositories. Eleven microarray datasets of human LECs, including 71 samples divided between control and cancer-related groups, were combined. Each study was first explored for potential inconsistencies and processed separately in a platform-specific manner (Affymetrix or Applied Biosystem) based on published methods. The in-house refined processing pipelines, implemented in R, included background correction, normalisation, summarisation, log2 transformation, filtering and gene annotation. The studies were merged, re-normalised, and batch effect corrected (comBat algorithm), removing the technical bias and favouring a clustering between the applied treatments. The batch effect removal was verified with principal component analysis and unsupervised clustering. Three different algorithms were used for network inference (ARACNE, GENIE3 and TIGRESS) with the final dataset containing 6456 genes for 71 samples as input. A sensitivity analysis was used for determining the network appropriate threshold needed to highlight the most relevant interactions. Only the top 15 % common interactions between the three algorithms were kept to build the final network composed of potential interactions between transcription factors and the genes they regulate. A combination of in silico and in vitro approaches was employed to validate the network’s major interactions. The inferred consensus network was analysed for topological consideration with graph theory metrics. Graph theory highlighted HEY1 as one of the most interconnected transcription factor in the top 10 of genes with the highest centrality values. The regulation pathway associated with this gene in the inferred network, HEY1-JUNB-HES1, was explored at the transcriptomic and protein levels using in vitro experiments (PCR and immunofluorescence). Interestingly, HEY1 and HES1 are already known to be involved in the NOTCH1 signalling pathway, which plays a role in tumour angiogenesis. It was an additional reason to investigate the regulations between JUNB, HEY1 and HES1 in a lymphangiogenic context. The biological validation experiments revealed that the expression of JUNB in LECs was dependent on the activation/inhibition of the NOTCH1 signalling pathway, and therefore of the HEY1-HES1 genes. At both transcriptomic and protein levels, JUNB was more (resp. less) expressed when the NOTCH1 signalling pathway was promoted (resp. inhibited). These findings were consistent with the regulation of HEY1, HES1 and JUNB highlighted in the inferred network. This work presents a dual in vitro-in silico approach to investigate the transcriptomic regulations in tumour lymphangiogenesis and proves the effectiveness of such integrative methods to uncover new insights in biological processes. Moreover, the development of new tools and pipelines to analyse already available biological data is an innovative and necessary way to better understand emerging collective behaviours from individual samples/studies.