RNA-Seq Expression Profiling of AML Stem Cells Reveals Differential Expression of Lineage Differentiation Markers and Novel Splice Variants.

Abstract 2502

Leukemia stem cells (LSCs) in acute myeloid leukemia (AML) are resistant to conventional chemotherapy, able to sustain leukemia through treatment, and may regenerate disease after treatment completion. Understanding the biological differences between LSCs and non-LSC leukemic blasts will be important for developing improved and targeted AML therapies. We report here the results of an RNA-Seq analysis of differential gene and splice variant expression in LSC-enriched vs. non-enriched leukemia fractions in 8 AML patients. Using fluorescence activated cell sorting (FACS) we divided 8 adult AML patient blast samples into LSC-enriched (“LSC” = Lin-, CD34+, CD38-) and “non-LSC” fractions. We extracted total RNA from each sample, purified mRNA, and performed massively parallel RNA sequencing using the Illumina HiSeq platform. Hierarchical clustering based on RPKM gene expression values showed greater similarity between LSC and non-LSC fractions within each patient's leukemia than similarity of LSC and non-LSC fractions among patients. Three hundred seventy-one genes were differentially expressed between the LSC and non-LSC fractions, with the most prominent public differences being increased expression of myeloid differentiation markers (e.g. LYZ, MPO, HBA/B) in the non-LSC samples. The LSC fractions displayed increased expression of CD34, FLT3, N-ras, K-ras, AML1, cyclin D1, CD61, and M-SCF. Pathway analysis revealed statistically significant differences in gene expression in the KEGG-defined Acute Myeloid Leukemia, Cell Cycle, and Hematopoietic Cell Lineage pathways between the LSC and non-LSC fractions. Using the MapSplice software package we identified multiple novel splice variants present in ≥ 6 of the 8 AML patient samples. These splices were found in genes known to be associated with leukemogenesis including MYC, MTOR, and FLT3. In addition to the identification of novel splices in AML associated genes, we used the FDM software package to measure differential splice utilization between the LSC vs. non-LSC fractions. Using FDM, we measured statistically significant splice utilization between LSCs and non-LSCs in 87 genes including ZNF34, NUMA1, NBEAL2, BAT2D1, FGR, MAP3K6, RFX5, NFATC2IP, and LRMP. In addition to these globally differentially spliced genes, several splice variants were highly differentially expressed between LSC and non-LSC fractions in individual patients (e.g. MAGED4, CYBB, IL17RA). Somewhat unexpectedly, no genes were both significantly differentially expressed and differentially spliced in the LSC vs. non-LSC fractions. To our knowledge, the data reported here represent the first RNA-Seq analysis of LSC-enriched vs. non-enriched AML fractions. Our results confirm prior gene-expression microarray studies that showed differential gene expression in LSC-enriched vs. non-enriched fractions. We extend this work by reporting the discovery of novel splice variants in AML associated genes and observe, for the first time differential splice utilization between LSC and non-LSC fractions. These results should prove valuable for both understanding LSC biology and guiding development of LSC-specific therapeutic strategies.