Deciphering Molecular Heterogeneity in Pediatric AML Using a Cancer Vs Normal Transcriptomic Approach

Introduction and Aim Although cytogenetics and response-guided therapy have considerably improved prognostication of pediatric AML (pedAML) patients, still 30-40% of the good responders relapse. Further delineation of the transcriptome of AML subpopulations, e.g. leukemic stem cells (LSCs), might result in a better understanding of pedAML biology and provide novel biomarkers for diagnostics, risk stratification, follow-up and targeted therapy. Methods Fluorescence-activated cell sorting (FACS) was used to isolate CD34+CD38- and CD34+CD38+ cells from pedAML patients/healthy controls (cord blood (CB), normal bone marrow (NBM)), defined as LSC/hematopoietic stem cell (HSC) and leukemic blast (L-blast)/control blast (C-blast), respectively. Sorting multiple phenotypes, both BM and blood, yielded 42 LSC and 35 L-blast fractions. Gene expression profiles (GEP) of LSCs and L-blasts were identified by a Cancer vs Normal (CvN) approach, whereas paired analysis of LSC vs L-blast aimed to identify LSC-specific aberrations. Micro-array analysis (4 pedAML, 3 CB) was followed by targeted quantitative PCR (qPCR) validation of the highest differentially expressed genes (DEGs) in LSC (n=52), L-blast (n=42) and between LSC and L-blast (n=15) (25 pedAML, 11 CB/9 NBM). DEGs were functionally analysed by protein association (STRING) and by gene set enrichment analysis (GSEA-Cytoscape). An overview of the workflow is shown in Fig. 1A. Results LSC vs HSC micro-array analysis revealed 83 up- and 212 downregulated targets (Fig. 1B). qPCR confirmed 8 and 11 of the 52 tested targets to be highly significantly up- and downregulated, respectively, in LSC (n=42) compared to HSC (n=20) (P L-blasts showed 157 and 332 up- and downregulated DEGs (Fig. 1C), and 8/42 were confirmed as significantly upregulated by qPCR (L-blast=35 vs C-blast=19, P Gene sets enriched in LSCs vs L-blasts addressed inflammatory responses, adipogenesis, TNF signaling and response. Pathway analysis showed repression of cell cycle genes, consistent with LSC quiescence. qPCR validation of 15 out of the 117 upregulated targets (Fig. 1D), of which 5/15 had a neural link (PCDHB2, GPRIN3, SLC22A23, CDR1, RPGRIP1L), did not confirm significant dysregulated expression (P>.05). Interestingly, the set of DEGs between LSCs and L-blasts shared only few genes with those differentially expressed between HSC and C-blast (28/306; 4 up- and 24 downregulated). This low intersection (7.7%) is in strong contrast to the previously reported 34% in adult AML (Gal et al. 2006). Moreover, none of those genes were detected in adult AML (except for CD38 downregulation). Remarkably, 3/4 mutual upregulated genes (IGF2, GPRIN3, PROS1, CYTH4) are involved in neural crossroads and have no relation to stemness nor AML. Conclusion Combining FACS with CvN transcriptomic profiling, followed by qPCR validation, identified novel DEGs in pedAML subpopulations. The majority of the most significant upregulated targets in LSCs (n=8) and L-blasts (n=8) showed no previous link to pedAML. Identification of 11 novel downregulated targets, often described as TSGs in solid tumors, warrants further studies whether hypomethylating therapy could result into LSC eradication in pedAML. LSCs reflected low cell cycle activity and elevated inflammatory response, hypoxia, metabolic dysregulation and signaling compared to HSC. GEP of LSCs vs L-blasts revealed a distinct molecular landscape compared to adult AML, and suggested a possible link between distortion of neural and hematopoietic systems in leukemogenesis. Figure Disclosures No relevant conflicts of interest to declare.