Identification of glucocorticoid-regulated genes and inferring their network focused on the glucocorticoid receptor in childhood leukaemia, based on microarray data and pathway databases

Abstract

Acute lymphoblastic leukaemia (ALL) has the highest mortality rate in childhood cancer. Glucocorticoids (GCs) have been used as chemotherapeutic drugs for children with ALL for more than 50 years. GCs induce apoptosis in lymphoid cells. However, little is known about the molecular mechanism of GC-induced apoptosis and there are many controversial hypotheses about genes regulated by GCs and their gene networks. In particular, two main issues are investigated: (i) GC-regulated genes and (ii) the glucocorticoid receptor (GR) gene networks. Only few overlapping genes have been reported from previous studies. Moreover, GCs function by binding with their receptors. The underlying mechanisms of cell type specific GR gene networks are not well established.

The goal of this thesis is to understand the mechanism of the GC-induced apoptosis mechanism. The first part of this thesis presents an identification of GC-regulated genes. This study uses secondary microarray data, originating from prednisolone (glucocorticoid) treated childhood ALL samples (Schmidt et al., 2006) (B-linage and T-linage) that were collected before treatment and at six and twenty four hours after treatment. We replicate the authors’ original study and discover more probe sets including all the probe sets from that original study. This result shows the robustness of this data. Then, we extend the data analysis and propose new criteria based on differences between T- and B-ALL patients. The results reveal the proposed GC-regulated genes. These candidate genes are grouped in order to find similar expression patterns which lead to possible co-regulated genes, or similar function and sharing networks and pathways. Four emergent clustering methods are used: Self organising maps (SOM), Emergent self organising maps (ESOM), the Short Time series Expression Miner (STEM) and Fuzzy clustering by Local Approximation of MEmbership (FLAME). These genes are used in the following gene expression analysis step.

The second part of this thesis focuses on inferring gene networks of GC-regulated genes and GR. There are many tools available for inferring gene networks including mathematical modelling and statistical methods. Each tool has its own advantages and disadvantages. For a modelling method, how do we know that the model represents the true relationship or interaction among genes? The need to verify results from modelling still exists. Prior knowledge has been used for this purpose. In this study, we use literature knowledge-based network tools, mainly the Ingenuity Pathway Analysis software (IPA) to elucidate gene networks. First, we illustrated gene networks at three time intervals and identified the prominent genes during those time points. Second, we further elucidated GR gene networks using gene lists from STEM. Third, we investigated the behaviour of selected known genes from the apoptosis, p53 and NFB pathways and inferred gene networks from the selected genes. Fourth, we inferred GR gene networks using the same gene list from previous studies (Phillip et al., 2005). We also used another two network tools: the BiblioSphere Pathway Edition (BSPE), and Oncomine to enhance the reliability of the gene network. Finally, we propose a GR gene network.

In summary, we undertook a gene to gene network of GC-induced apoptosis process based on childhood leukaemia patients. This study identified novel genes and their functions, and pinpointed possible gene networks which provide information for future research.