In this study, we presented a detailed, multi-stage, translational bioinformatic analysis of ApoE knockout mice, exploiting different methods in order to identify critical altered molecular mechanisms and important central players. Our approach was to apply a generic computational framework, which exploits rigorous statistical or computational measures at every analytical step, for the efficient systems level interpretation of the results of ApoE dataset. The workflow proposed here integrates various software products, in a unified translational pipeline, able to cope with versatile, high-volume investigation tasks, and at the same time provide a reliable systemic interpretation for the biological mechanism studied. In this way, a powerful translational backbone is set, which connects the wet-lab part with the theoretical knowledge for the biological problem interrogated, as rescued in molecular databases, controlled ontological vocabularies or the literature. The workflow presented in this study, currently in the phase of implementation as regards to its software components integration, represents an efficient and highly innovative effort, either in terms of speed of analytical performance, as well as real biological value of the results. This is so because it provides results which are qualified from a composite framework that combines ideally both individual and group quality measures, together with an insightful comprehension of the underlying topological networks, actively involved in the mechanism studied. The correlation of the results of the molecular analysis with literature-derived associations manages to highlight and propose promising, novel candidates that have not been studied in the context of the given pathology. They could thus represent ideal targets for further biological experimentation. Maximizing the total information gain encompassing all analytical steps of the proposed workflow represents a critical parameter regarding the implementation of the web application. However as the derivation of automated statistical thresholds for such high-volume data processing in an unsupervised manner both in terms of performance and computational speed is a very challenging task, this still remains an open issue for extensive research work and testing, representing an important point for future work.
Computational technologies are complementary to conventional “wet lab” gene discovery technologies in that they can support the prioritization and comprehension of high-volume molecular data (i.e., omic datasets from a microarray or novel sequencing technologies, associated regions from genome wide association or linkage studies) enabling the efficient selection of the top candidates, under a range of criteria, for further study. In recent years, there are popular tools and environments in the scientific computing realm (data-mining, artificial intelligence, hyper-computing), like the Taverna workflow manager [24
] or the RapidMiner solution, formerly known as YALE [25
], which enable efficient workflow integration and deployment, exploiting versatile web service repositories, containing hundreds of operators implementing various analytical tasks. Especially Taverna workflow manager, through myExperiment (http://www.myexperiment.org
] and Biocatalogue (http://www.biocatalogue.org
], integrates an impressive number of workflows and web services all accessible through Taverna, for a very wide range of disparate bioinformatics tasks. However, to the best of our knowledge, the workflow showcased in this work addresses in a sequential, unsupervised fashion disparate tasks enabling and empowering decisively the translational procedure, in a completely innovative yet efficient way.
Applying the proposed workflow to a dataset from ApoE knockout and wild type mice, it was shown that the gene expression profile in atherosclerotic plaques containing arteries of ApoE knockout mice is profoundly different from wild type. Specifically, 852 genes were found as differentially expressed and the majority of them appear after the age of 32 weeks. The indicated altered processes, as revealed by ontology-based enrichment analysis, include adhesion and signal transduction, differentiation, apoptosis, and immune response, reflecting the cellular and molecular complexity of atherosclerosis and the cross-talk of endothelial and immune cells in aortic lesions. Cluster analysis revealed three major groups of genes with similar expression profiles, which were further analyzed, in order to find functional (GO-based) subgroups in each cluster. In agreement with the notion that atherosclerosis is an inflammatory disease [2
], immune response and inflammation were the prominent categories indicated as significantly altered in the case of cluster 2, which contains genes upregulated both in 36- and 78-weeks-old mice. Promoter analysis of the genes under these categories revealed common binding elements that could contribute to a common transcriptional regulation. In particular, all of the tested genes (100%) contain cis-elements of the KLF and RXR family. The KLF family groups binding sites recognized by Krueppel like transcription factors (KLFs) [28
] which are involved in many physiological and pathological processes, such as cell differentiation, proliferation, cell growth, and apoptosis during normal development or under different disease conditions. It is noteworthy that KLFs have been implicated in acute and chronic inflammatory disease states, such as atherosclerosis, diabetes, and airway inflammation [29
]. It is important to note that despite the identification of KLF cis
-acting elements, Klf4 TF was also found upregulated, suggesting that this factor could be involved in the regulation of the observed stimulation of the immune response related mechanisms. Klf4 has been found to regulate monocyte differentiation and to activate the macrophages to induce inflammation [30
]. Furthermore, Klf4 regulates the proliferation and differentiation in vascular smooth muscle cells after injury to the vessel [31
]. However, Klf4 seems to have an anti-inflammatory role in endothelial cells [28
]. Regarding RXR family, it groups together motifs related to the receptors of retinoids, which are recognized by various heterodimers of retinoid X receptors (RXRs) and retinoic acid receptors. Interestingly, RXR has been reported to regulate several genes related to metabolic homeostasis and inflammation [32
]. RXR form heterodimers with many different nuclear receptors, PPARs, LXR, and FXR affecting different aspects of cholesterol metabolism in macrophages, something known to be important in the development of atherosclerosis [32
]. In addition, among the identified putative TF binding sites there are interferon regulatory factors-related elements (IRFs) in the 86% of the promoters, as well as glucocorticoid responsive elements (GREs) in the 92% of the tested promoters. In agreement, Irf8, a transcription factor involved in modulation of immune response and as a central element in the IFN signaling cascade, was found significantly overexpressed, suggesting that Irf8, together with Klf4, could be involved in the upregulation of the immune response related genes. Regarding GREs, it is well known that glucocorticoid receptors play important roles in both physiological and pathological conditions involving immunity and inflammation and that they are involved in the pathology of cardiovascular diseases [33
]. Finally, includes several genes implicated to various aspects of the disease. It is noteworthy to mention Tlr2, a member of the Toll-like receptors family, which plays a fundamental role in activation of innate immunity [34
]. Furthermore, the identification of Psen2 (presenillin 2), a gene implicated in Alzheimer's disease, as candidate hub gene is interesting because genes implicated in Alzheimer's have been reported to affect cholesterol or lipoprotein function and have also been implicated in atherosclerosis [35
Concluding, this bioinformatic analysis of ApoE knockout mice revealed critical altered cellular mechanisms governing atherosclerosis and indicated important molecular players that could be important targets for treatment of this complex disease.