Gene Expression Profiling of CD93-Selected CP-CML Stem Cells Confirms Their Quiescent Character and Biomarker Potential

The introduction of BCR-ABL tyrosine kinase inhibitors has revolutionized the treatment of chronic myeloid leukemia (CML). A major clinical aim remains the identification and elimination of low-level disease persistence, termed "minimal residual disease". Disease persistence suggests, that despite targeted therapeutic approaches, BCR-ABL-independent mechanisms exist which sustain the survival of a small population of cells, termed leukemic stem cells (LSC).

We previously identified CD93 expression as a promising biomarker of LSC in chronic phase (CP)-CML. Our group has described the long term self-renewal potential of Lin^-CD34⁺93⁺ CP-CML cells compared to their Lin^-CD34⁺93^- counterparts through LTCIC assays (n=3, p<0.0001) and NSG engraftment models (3.5-30-fold increased in engraftment with Lin^-CD34⁺93⁺ cells, p<0.03). We hypothesized that CD93⁺-selected cells would represent a more immature functional phenotype compared to CD93^- selected cells. The aim of this study was to characterize differences in the gene expression profile between CD93⁺ and CD93^- CML LSC populations and determine heterogeneity of each population at a single cell level.

To interrogate this, we initially identified CP-CML subpopulations with the greatest functional capability compared to normal. Normal and CP-CML samples were FACS-sorted into HSC/LSC, CMP, GMP, and MEP sub-populations. Results suggest a significant change in functional status between normal and CP-CML subpopulations within the HSC/LSC compartment (lin^-CD34⁺CD38-CD45RA^-CD90⁺), where CML LSC demonstrated significantly increased proliferation (14 fold expansion; P<0.001) compared to normal HSC (no expansion) after 5 days in vitro culture in physiological growth factors. In addition, equivalent numbers of CML LSC produce ~4-fold more colonies in colony forming cell (CFC) assays than normal HSC (329±56 versus 86±17 per 2,000 cells, respectively (p<0.05)). Furthermore, fluorescence in situ hybridization demonstrated that >90% of lin^-CD34⁺CD38^-CD45RA^-CD90⁺ CML LSC from all patient samples were BCR-ABL positive. Subsequent experiments were confined to the LSC population.

We hypothesized that lin^-CD34⁺CD38^-CD90⁺CD93^- CML cells would have a more mature gene expression profile compared to lin^-CD34⁺CD38^-CD90⁺CD93⁺ cells. CP-CML cells were sorted into (1) lin^-CD34⁺, (2) lin^-CD34⁺CD38^-CD90⁺CD93^- and (3) lin^-CD34⁺CD38^-CD90⁺CD93⁺ populations. RNA was harvested at baseline from bulk populations (1) to (3) and cDNA was generated from single cells using the Fluidigm C1 autoprep system. Using Fluidigm technology, quantitative PCR of 90 lineage-specific and cell survival genes was performed within all populations of cells (1) to (3) in 'bulk' samples (n=3), and at single cell level (n=123 CD93⁺, n=120 CD93^-single cells; n=3 samples in total).

Bulk sample analysis demonstrated a significant increase in expression of lineage commitment genes within the lin^-CD34⁺CD38^-CD90⁺CD93^- population, as shown by increased expression of GATA1 (p=0.0007), and CBX8 (p=0.0002). The lin^-CD34⁺CD38^-CD90⁺CD93⁺ population displayed a less lineage-restricted profile with increased expression of CDK6 (p=0.05), HOXA6 (ns), CDKN1C (ns) and CKIT (p=0.0014), compared to the lin^-CD34⁺CD38^-CD90⁺CD93^- population. Furthermore, the two populations could be segregated by differential gene expression through gene clustering.

At a single cell level, differences were noted in the frequency of expression between lin^-CD34⁺CD38^-CD90⁺CD93^- and lin^-CD34⁺CD38^-CD90⁺CD93⁺ populations, particularly in GATA1, TPOR, and VWF. Although a statistically significant change was demonstrated in gene expression between the lin^-CD34⁺CD38^-CD90⁺CD93^- and lin^-CD34⁺CD38^-CD90⁺CD93⁺ populations in a number of genes, we were not able to segregate the populations by differential expression using gene clustering. This highlights the heterogeneous nature of the cell populations and the inability to distinctly characterize between the two populations at a single cell level.

Our results validate CD93 as a potential biomarker to separate the primitive CP-CML LSC population and highlight key lineage and cell survival pathways that are altered in CML LSC. The results demonstrate the heterogeneity seen within gene expression at the single cell level, which may allow for further insight into the CML LSC compartment with further analyses.