Identification of Genes Whose Expression Overlaps Age Boundaries and Correlates with Risk Groups in Paediatric and Adult Acute Myeloid Leukaemia

Abstract

Few studies have compared gene expression in paediatric and adult acute myeloid leukaemia (AML). In this study, we have analysed mRNA-sequencing data from two publicly accessible databases: (1) National Cancer Institute's Therapeutically Applicable Research to Generate Effective Treatments (NCI-TARGET), examining paediatric patients, and (2) The Cancer Genome Atlas (TCGA), examining adult patients with AML. With a particular focus on 144 known tumour antigens, we identified STEAP1, SAGE1, MORC4, SLC34A2 and CEACAM3 as significantly different in their expression between standard and low risk paediatric AML patient subgroups, as well as between poor and good, and intermediate and good risk adult AML patient subgroups. We found significant differences in event-free survival (EFS) in paediatric AML patients, when comparing standard and low risk subgroups, and quartile expression levels of BIRC5, MAGEF1, MELTF, STEAP1 and VGLL4. We found significant differences in EFS in adult AML patients when comparing intermediate and good, and poor and good risk adult AML patient subgroups and quartile expression levels of MORC4 and SAGE1, respectively. When examining Kyoto Encyclopedia of Genes and Genomes (KEGG) (2016) pathway data, we found that genes altered in AML were involved in key processes such as the evasion of apoptosis (BIRC5, WNT1) or the control of cell proliferation (SSX2IP, AML1-ETO). For the first time we have compared gene expression in paediatric AML patients with that of adult AML patients. This study provides unique insights into the differences and similarities in the gene expression that underlies AML, the genes that are significantly differently expressed between risk subgroups, and provides new insights into the molecular pathways involved in AML pathogenesis.

Keywords: HOX; SOX; WNT; antigen identification; mRNA-sequencing; meta-analysis; paediatric and adult acute myeloid leukaemia.

Figure 1

Overview of the analytical workflow and identification of differentially expressed genes (DEGs) that overlap between subgroup comparisons. (A) DESeq2 was used to perform differential gene expression (DGE) analysis using mRNA-Seq raw counts for various poor prognostic versus good prognostic group comparisons within Therapeutically Applicable Research to Generate Effective Treatments (TARGET), and The Cancer Genome Atlas (TCGA) patient datasets. DGE results were filtered (Benjamini–Hochberg (BH)-adjusted p value < 0.01) to yield significant DEGs. Where genes of interest (GOI) were found to be significantly differentially expressed, raw count data within a corresponding subgroup comparison was transformed and expression levels of GOI across all patients in the comparison were assigned to a quartile (Q1/Q2/Q3/Q4). Log-rank tests were carried out between patients of Q1 and Q4 expression, and of Q2 and Q3, with p value < 0.05 deemed significant. Pathway analysis was also carried out using the Kyoto Encyclopedia of Genes and Genomes (KEGG) 2016 Database on total significantly differentially expressed GOI, on those considered as downregulated (log2 fold change < 0), and those considered as upregulated (log2 fold change > 0); (B) Venn diagrams provide a visual summary of (i) all genes and (ii) GOI that were identified as significantly differentially expressed when comparing TARGET and TCGA subgroups.

Figure 2

DGE analysis. Pairwise comparisons of event-free survival (EFS) of (A) TARGET patients by standard versus low risk (p = 0.02), and (B) TCGA patients by (i) intermediate versus good risk (p = 0.01) and (ii) poor versus good risk (p = 0.007) subgroups. All p values derived from log-rank analysis. Volcano plots represent the genes that were differentially expressed by p > 0.01 (grey dots) or p < 0.01 (green dots) when comparing (C) TARGET patients by standard versus low risk, and (D) TCGA patients by (i) intermediate versus good and (ii) poor versus good risk subgroups. Names and black dots indicate the position of antigens listed as GOI when studying gene expression.

Figure 3

Significant impacts of gene expression levels (divided into quartiles) on patient EFS. (A) TARGET study patients across the standard and low risk subgroups were found to have significant differences in EFS when comparing quartile expression levels of (i) BIRC5; (ii) MAGEF1; (iii) MELTF; (iv) STEAP1 and (v) VGLL4; (B) TCGA study patients across the intermediate and good risk subgroups showed significant differences in EFS when comparing quartile expression levels of (i) MORC4, and patients across poor and good risk subgroups showed significant differences in EFS when comparing quartile expression levels of (ii) SAGE1.