Bi-Phenotypic Leukemic Cells as Transient Intermediates Between Leukemic Stem Cells with Lymphoid Characteristics and the Leukemic Myeloid Bulk: Insights from a Murine Model of CALM/AF10 Positive Acute Myeloid Leukemia

Co-expression of lymphoid and myeloid surface markers is a common phenomenon in human acute myeloid leukemia (AML). One explanation for this could be the aberrantly expression of lymphoid antigens on malignant myeloid cells. Another intriguing possibility is that these leukemic bi-phenotypic cells emerge from a leukemic stem cell (LSC) with lymphoid characteristics. In this scenario co-expression of lymphoid and myeloid markers on a substantial proportion of the leukemic cell population would rather reflect the nature of the LSC in these AML cases than deregulated antigen expression of a malignant transformed cell. We have recently characterized the LSCs in a murine model of CALM/AF10 positive AML. In each leukemic mouse we could demonstrate the presence of three distinct populations, a minor B220⁺/myeloid marker (MM)⁻ population which showed the highest leukemogenicity in vitro and in vivo (candidate LSCs), a bi-phenotypic population which was MM⁺/B220⁺, and the leukemic bulk population which expressed only myeloid markers without the expression of the lymphoid associated marker B220 (MM⁺/B220⁻) (Deshpande et al. Cancer Cell 2006). We now sought to analyze the relationship between these three leukemic populations. Single cell-sorted LSC candidates (B220⁺/MM⁻) could give rise to the other leukemic populations in vitro. Real time RT-PCR for 5 lymphoid-associated and 5 myeloid-associated transcripts showed that most lymphoid-associated genes were highly expressed in the B220⁺/MM⁻ cells and their expression levels were decreased or not detectable in the myeloid bulk. The converse was true with the myeloid associated transcripts confirming that the LSCs were more lymphoid-like than the myeloid bulk that they generate. The expression of lymphoid-associated genes such as Rag2, Pax5, IL7Ra and Ebf1 showed a progressive downregulation and that of myeloid associated transcripts such as c-fms, GM-CSFR, GCSF-R showed a progressive upregulation upon differentiation of single B220 positive cell to the myeloid lineage. The expression of most (7 out of 10) of these lineage-associated genes was intermediate in the B+/MM+ cells suggesting a transient lympho-myeloid stage of differentiation. We then performed a DNA microarray analysis of these sub-populations by using a chip with 20,172 PCR-amplified, sequence-verified, gene-specific DNA fragments (LION Biosciences). Correlation based principal component analysis (PCA) was conducted on the normalized expression values of genes showing differential expression in the three cell populations. Two statistically significant principal components were extracted, explaining more than 95% of the variation, where the genes were clustered into three classes based on their loadings. These clusters were overlaid with the loadings of the three cell populations to see whether the clusters match to a certain cell type. We observed that the three cell populations actually corresponded to the three clusters of genes and interestingly the B220⁺/MM⁺ bi-phenotypic cells lie between the B220⁺/MM⁻ LSCs and the B220⁻/MM⁺ bulk. These data demonstrate that the AML in the CALM/AF10 murine model is arranged as a hierarchy of progressively differentiated cells with the B220⁺/MM⁺ bi-phenotypic cells as transient intermediates between the B220⁺/MM⁻ LSCs and the myeloid B220⁻/MM⁺ bulk. These data support a model in which myelo-lymphoid cells observed in cases of human AML could arise as intermediates during the differentiation process of LSCs with lymphoid characteristics into leukemia with predominantly myeloid features. These observations encourage studies which test the hypothesis that AML subtypes characterized by co-expression of lymphoid and myeloid surface markers are originating from a leukemia propagating cell with lymphoid characteristics. The presence of lymphoid markers on LSCs in certain AML sub-sets could pave the way for a more effective anti-leukemic therapy.