Meta-analysis of gene expression in relapsed childhood B-acute lymphoblastic leukemia

Abstract

Background: Relapsed pediatric B-acute lymphoblastic leukemia (B-ALL) remains as the leading cause of cancer death among children. Other than stem cell transplantation and intensified chemotherapy, no other improved treatment strategies have been approved clinically. Gene expression profiling represents a powerful approach to identify potential biomarkers and new therapeutic targets for various diseases including leukemias. However, inadequate sample size in many individual experiments has failed to provide adequate study power to yield translatable findings. With the hope of getting new insights into the biological mechanisms underpinning relapsed ALL and identifying more promising biomarkers or therapeutic targets, we conducted a meta-analysis of gene expression studies involving ALL from 3 separate studies.

Method: By using the keywords "acute lymphoblastic leukemia", and "microarray", a total of 280 and 275 microarray datasets were found listed in Gene Expression Omnibus database GEO and ArrayExpress database respectively. Further manual inspection found that only three studies (GSE18497, GSE28460, GSE3910) were focused on gene expression profiling of paired diagnosis-relapsed pediatric B-ALL. These three datasets which comprised of a total of 108 matched diagnosis-relapsed pediatric B-ALL samples were then included for this meta-analysis using RankProd approach.

Results: Our analysis identified a total of 1795 upregulated probes which corresponded to 1527 genes (pfp < 0.01; FC > 1), and 1493 downregulated probes which corresponded to 1214 genes (pfp < 0.01; FC < 1) respectively. S100A8 appeared as the top most overexpressed gene (pfp < 0.01, FC = 1.8) and is a potential target for further validation. Based on gene ontology biological process annotation, the upregulated genes were most enriched in cell cycle processes (enrichment score = 15.3), whilst the downregulated genes were clustered in transcription regulation (enrichment score = 12.6). Elevated expression of cell cycle regulators (e.g kinesins, AURKA, CDKs) was the key genetic defect implicated in relapsed ALL, and serve as attractive targets for therapeutic intervention.

Conclusion: We identified S100A8 as the most overexpressed gene, and the cell cycle pathway as the most promising biomarker and therapeutic target for relapsed childhood B-ALL. The validity of the results warrants further investigation.

Keywords: Gene expression; Microarray; Pediatric B-acute lymphoblastic leukemia; Relapse.

Fig. 1

Venn diagram of differentially expressed probes identified from each individual microarray dataset using limma approach. a Upregulated probes (p-value < 0.01, logFC > 1); b Downregulated probes (p-value < 0.01, logFC < -1). Only 2 probes which encode for LTF and S100A8 were found concordantly upregulated in 2/3 studies

Fig. 2

Heatmap of the top 100 differentially expressed probes between relapsed and matched diagnosed B-ALL samples (n = 108) from meta-analysis of three microarray datasets. Each green color column denotes newly diagnosed B-ALL samples whilst each blue color column denotes relapse B-ALL samples. Expression levels are represented by red (high expression) and green (low expression)

Fig. 3

The ten most significant biological processes associated with genes upregulated in relapsed childhood B-ALL

Fig. 4

The ten most significant biological processes associated with genes downregulated in relapsed childhood B-ALL

Fig. 5

Protein-protein interaction network of cell cycle genes identified in top 100 upregulated probes in relapsed childhood B-ALL