Functional ABCG2 Is Expressed on CML Stem Cells and Its Inhibition Selectively Depletes CML CD34+ Cells.

We have previously described a population of deeply, but reversibly, quiescent stem cells (qSC) found in patients with chronic phase (CP) CML at diagnosis. In vitro studies have proven this population to be highly insensitive to imatinib mesylate (IM; Gleevec, STI571) induced killing, and more worryingly shown that qSC are accumulated after CML CD34+ cells are treated with IM. As it is likely that CML qSC closely resemble normal HSC, we hypothesise that they too may express the stem cell-associated ABCG2 and have therefore examined the expression and function of this drug efflux pump on CML cells. In agreement with other studies we show the interaction between ABCG2 and IM. Using ABCG2 over-expressing cells (AML6.2 and HL60-BCRP) we found that ≥0.5μM IM reduced efflux of the ABCG2 substrate BODIPY-Prazosin by a similar degree as the inhibitor fumitremorgin C (FTC; 10μM). We have now examined expression and function of ABCG2 on primary CML cells taken from patients in chronic phase (CP) and prior to any treatment. Quantitative Taqman analysis of 8 CD34+ enriched (≥90%+) CML samples revealed that the level of expression is 2.46 fold higher than that in normal mobilised CD34+ cells (n=8 CML, n=4 normal). In addition, we undertook microarray analysis of normal or CML CP CD34+ cells fractionated according to cell cycle using Hoechst-Pyronin (G0, G1 and G2/S/M). These analyses (n=3 normal, n=5 CML) show that at all stages of the cycle CML cells express more ABCG2 than normal cells and that G0 CML cells express 2.48 fold more than those in G1 , confirming both the over-expression in CML and relationship to the most primitive subset of cells. Using the antibody BXP21 we found that 8 of 9 samples contain ABCG2+ve cells (5 of 9 ≥60% of cells ABCG2+). We also examined the function of ABCG2 on CML CD34+ cells by performing efflux assays, 4 of 6 showed efflux that was inhibited by 10μM FTC or ≥0.5μM IM, and this efflux capacity correlated with BXP21 staining. We therefore considered whether the combination of IM therapy and ABCG2 inhibition would overcome the accumulation of CML qSCs we have previously reported after treatment with IM. Using CFSE to track cell division we treated CD34+ enriched CML samples with 5μM IM +/− FTC or with 10μM FTC alone for 3 days. In comparison to untreated controls 5μM IM reduced the total number of cells to 31.9±9.2 % and the number of CD34+ cells to 43.2±17.6%. However, the non-cycling qSC significantly increased to 318±75.8% of control. In contrast, the ABCG2 inhibitor FTC did not effect a reduction in total cells (99.5±11.9%) but gave a significant reduction of CD34+ cells (58.6±8.4%; p=0.02) and no accumulation of qSC (104.6±33.8%) when used alone. We saw no cumulative effect when IM and FTC were given concurrently. These data suggest strongly that FTC may be used to deplete CD34+ ‘stem cells’ from CML, as the total cell number is unchanged it is likely that this depletion is by the induction of differentiation. We propose that the expression of ABCG2 may be clinically significant in CP CML and that inhibition of this pump may result in a ‘stem cell targeted therapy’ that could be followed by IM treatment to reduce the tumor load. Such reduction of CML stem cells would result in elimination of minimal residual disease and effect a lasting remission.