Hematopoietic stem cells (HSCs) are pluripotent stem cells that give rise to all the circulating blood cell types. They are defined by their ability to self-renew while generating differentiated daughter cells [9]. The maturation of HSCs into blood cells has been suggested to follow a hierarchical model (Fig. 14.2), in which the stem cells undergo a reduction in self-renewal potential while simultaneously gaining functional specialization [10]. Early on, HSCs make a fundamental commitment to either the lymphoid or myeloid fate. These two precursor types subsequently undergo further restriction in developmental potential until terminally differentiated cells are generated [11]. The common myeloid progenitors give rise to monocytes, neutrophils, basophils, eosinophils, erythrocytes, or megakaryocytes, while the common lymphoid progenitors give rise to B or T cells [12].

Cancer is a clonal disease that initiates in a single cell that has accumulated sufficient genetic damage to cause uncontrolled proliferation, and therefore the progeny of this cell makes up the tumor [13]. The experimental setting of transplantation, where subsets of tumor cells are transferred from one organism to another, demonstrates that tumors are ­functionally heterogeneous, as only a certain subpopulation of cells has the capacity to maintain or reinitiate the tumor. It is not always easy to define which cells within the population have this capability for self-renewal or tumor re-initiation.

Two general, but distinct models have been proposed to account for the heterogeneous proliferative potential of tumor cells (Fig. 14.3). The stochastic model proposes that all cells within the tumor are equivalent with respect to their potential to re-initiate the clone, however, any one cell would have a low probability of exhibiting this potential. In contrast, the hierarchical model predicts that re-initiation potential is a property of only a distinct subset of cells within the tumor, the so-called “cancer stem cell” [14].

There is strong support for the idea that cancer, and in particular leukemia, is a stem cell disease. Persuasive evidence for this concept comes from the discovery that most AML blasts do not proliferate [15, 16] and that only 1–4 % of leukemic cells transplanted in vivo are able to form spleen colonies. The most compelling evidence for a leukemia stem cell (LSC), or leukemia initiating cell (LIC), has been provided by a Canadian group based at the University of Toronto. In a seminal paper in this field, Lapidot et al. [17] reported that xenograft transplantation of only a few, highly selected human LSCs into immunodeficient mice was able to recapitulate a leukemic disease that reproduced many features of the AML seen in the original human patients. Subsequently, cell purification based on cell surface markers demonstrated that the cells capable of initiating human AML in non-obese diabetic mice with severe combined immunodeficiency disease (NOD/SCID mice) were from a rare subpopulation of CD34⁺/CD38⁻ cells. Normal HSCs share this same CD34⁺/CD38⁻ phenotype, suggesting that normal primitive cells, rather than committed progenitor cells, are the target for malignant transformation.

Further studies support a multistep process whereby an initial molecular aberration occurs at the level of the stem cell, creating a pre-leukemic LSC. Either these cells or their progenitors then undergo additional oncogenic events that ultimately lead to transformation into overt disease [18, 19]. LSCs themselves also demonstrate self-renewal heterogeneity, with a population of quiescent stem cells capable of endless self-renewal, progenitor populations with more limited potential for self-renewal, and, finally, a population of blasts with no self-renewal potential that is responsible for the clinical manifestations of the disease [20]. This hierarchical arrangement of LSCs mirrors that of HSCs and lends credence to the idea that, in leukemia, the primary target for cell transformation resides within the HSC population.

The concept of the LSC has increasingly important implications for treatment of acute leukemia. Most conventional chemotherapies and even existing targeted therapies may fail to affect the LSC. This may be attributable to the fact that the majority of the LSC population exists in the G0 phase of the cell cycle, where the absence of DNA synthesis, cell division, or any activity related to self-renewal, renders it insensitive to anti-leukemic drugs. The induction of LSCs to enter the cell cycle, where they are then vulnerable to destruction by standard chemotherapy drugs, represents one therapeutic strategy. Indeed, Saito [21] and colleagues demonstrated in a mouse model of human AML that treatment with granulocyte colony-stimulating factor (G-CSF), a cytokine that induces cell cycle entry of hematopoietic stem cells, caused the transplanted leukemia stem cells to proliferate and rendered them susceptible to the chemotherapeutic cytarabine. This finding provides proof of principle that the eradication of leukemia at its roots can be accomplished by developing treatments focused on pushing LSCs out of their quiescent state. An analogous approach, in which newly diagnosed AML patients received chemotherapy, with or without G-CSF, showed the addition of G-CSF concurrently with chemotherapy improved disease-free survival [22].

Additionally, HSCs interact with their microenvironment, termed the “stem cell niche,” through interactions between the chemokine SDF-1 and its receptor CXCR-4, and by the adhesion molecules VCAM/VLA4 and Angiopoietin-1/TIE-2. The same molecules are also involved in leukemia stem cell–niche interactions, and elevated levels of CXCR4 and VLA4 have been associated with poor response to chemotherapy and unfavorable prognosis in AML. Therefore, it may be possible to affect leukemia stem cells directly by targeting the specific molecules mediating LSC–niche interactions. Finally, in addition to nonspecific treatments, such as G-CSF, that manipulate LSC behavior, targeting LSC-specific molecules may represent an effective therapy for AML. For example, specific targeting of the PML–RARA leukemic oncoprotein expressed in LSCs with arsenic trioxide may be the mechanism by which this treatment achieves long-term remission of murine, and possibly human, leukemia [23].

Although not all malignancies fit into the cancer stem cell model, AML in particular is one of the diseases for which there is solid evidence of stem cell-like behavior. Consequently, there is considerable leukemia research devoted to further understanding the unique biologic and molecular properties of LSCs, by comparison with both their non-self-renewing downstream progeny and their normal HSC counterparts.

Epigenetic regulation refers to those heritable changes in gene function that occur without alteration of the primary DNA sequence [24]. The remodeling of chromatin, through DNA methylation and/or posttranslational modification of nucleosomal histones, has been implicated in the regulation of transcription. Anomalous epigenetic changes occur ­frequently in acute leukemia. The recurring chromosomal translocations seen in AML and ALL result in the generation of chimeric fusion proteins, as discussed in more detail later. Several of these fusion oncoproteins contribute to the development of leukemia partly by disrupting the modification of chromatin, through recruitment of chromatin-modifying coregulators [25]. The emerging link between the enzymes that modify chromatin and the transcription factors that control leukemic cell development and function strongly suggests a prominent role for epigenetic regulation in establishing and maintaining leukemic gene expression programs.

A new and exciting avenue of research in the molecular origins of leukemia has been launched with the discovery of microRNAs (miRNAs). MicroRNAs are endogenous ∼22 nucleotide noncoding RNAs that function as posttranscriptional silencers. They bind to complementary sequences present in coding mRNAs, usually in the 3′ untranslated regions, resulting in the degradation or interference with translation of the target mRNA (Fig. 14.5). These posttranscriptional regulators were inadvertently discovered in 1993 by Rosalind Lee, Rhonda Feinbaum, and Victor Ambros during the analysis of a Caenorhabditis elegans mutant [42]. MicroRNAs are predicted to target approximately 60 % of human protein-coding mRNAs [43] and are present in all human tissues [44]. A single miRNA can downregulate the production of hundreds of proteins [45]. miRNAs play an important regulatory role in many crucial physiological processes such as embryonic development, cellular apoptosis, and proliferation. Emerging evidence relates miRNA expression to various types of cancer, where crucial miRNAs have been shown to be either upregulated or downregulated. miRNAs function in complex regulatory networks to regulate hematopoietic differentiation and there seems to be an important correlation between the deregulation of miRNA expression and the presence of well-characterized leukemic cell abnormalities.

Some miRNAs have been shown to function as oncogenes and their overexpression plays a part in leukemogenesis. For example, the miR-17-92 cluster encodes for seven individual miRNAs that are frequently overexpressed in MLL-rearranged AMLs [46] and ALLs [47]. The forced expression of MiR-17-92 inhibits apoptosis and increases cell viability. More importantly, the expression of this miRNA alone, and in combination with an MLL fusion, enhances the proliferation while inhibiting the differentiation of hematopoietic progenitor cells. Thus, aberrant overexpression of miR-17-92 in MLL-rearranged acute leukemias could play a role in the development of MLL-rearranged leukemias. MiR-155 has also been identified as a significant miRNA expressed by hematopoietic cells. Its expression is tightly regulated in myeloid and lymphoid cells and increases after exposure to inflammatory stimuli. MiR-155 has been implicated in hematopoietic disease in three ways: first, MiR-155 is ­overexpressed in patients with AML-M4 and AML-M5 (according to the French–American–British classification system of acute leukemia); second, its sustained overexpression in mouse HSCs results in a myeloproliferative disorder; and third, it has been shown to directly repress key mRNAs involved in hematopoiesis [48]. Finally, miR-196a and miR-196b are upregulated in MLL-associated AMLs and AMLs with nucleophosmin (NPM) mutations. The expression of a MLL fusion protein caused overexpression of miR-196b and knockdown of the miRNA abrogated the replating potential of bone marrow cells expressing the MLL fusion [49]. Again, forced expression of miR-196b in bone marrow cells increased proliferative capacity and survival and led to a partial block in differentiation. The common leukemogenic mechanism of action of all three oncogenic miRNAs seems to be through increasing proliferation while blocking differentiation of hematopoietic progenitor cells.

MiRNAs are also hypothesized to function as tumor suppressors and may be underexpressed in certain acute leukemias. The best characterized in the context of acute leukemia is the let-7 family, which negatively regulates the human RAS [50] and HMGA2 [51] oncogenes. Its expression is found to be decreased in core-binding factor leukemias [52] (see later section on these leukemias) and is increased in APL cells when induced to differentiate with pharmacological doses of all-trans retinoic acid [53].

There are still unresolved paradoxes concerning miRNAs and their role in promoting acute leukemia. Computational and experimental analyses suggest that each miRNA regulates hundreds of targets [45], yet genetic studies usually show that the pleiotropic phenotypes that result from the loss of a miRNA can often be suppressed by the mutation of just one particular target gene [54]. The surprisingly modest extent of miRNA-mediated target repression (which rarely exceeds twofold [45]) seems inconsequential when compared to other normal intrinsic variations in gene expression [55]. One view is that miRNAs may act as a rheostat and fine tune the expression of most of their targets. This may be particularly relevant in the development of leukemia, as even a fine-tuning of the expression of a particular protein may be sufficient to disrupt the delicate balance between proliferation and differentiation.

The amount of an active protein available to carry out defined physiological functions is ultimately what governs cell growth and proliferation. Protein translation represents one important level of regulation of protein levels in cells, and protein synthesis is coupled to essential cellular events such as growth, proliferation, differentiation, and apoptosis. In normal hematopoietic cells, messenger RNA (mRNA) ­translation is tightly regulated at the initiation phase by ­signal transduction pathways that modify two proteins, ribosomal protein S6 kinase (S6K1) and eukaryotic translation initiation factor 4E (eIF4E). When activated, both of these proteins bind at the 5′untranslated region (5′ UTR) of mRNA and enhance the binding of mRNA to polysomes, leading to increased translation.

Deregulated translational control can therefore serve as a driving force for the oncogenic process. Indeed, morphological changes of the nucleolus, the site of ribosomal RNA synthesis, have been recognized as a consistent feature of cancer cells since the end of the nineteenth century [56]. mRNA translation is an extremely complex process in the eukaryotic cell and deregulation leading to oncogenesis could occur at numerous levels. Three hypotheses speculating on the mechanism by which aberrant mRNA translation could promote neoplastic transformation have been put forward [57]:

In terms of leukemogenesis, current evidence supports the mechanism suggested in the second scenario. eIF4E is a proto-oncogene that enhances both mRNA export from the nucleus and mRNA translation in the cytoplasm of a subset of transcripts coding for proteins involved in cell survival and proliferation, thereby increasing their availability to the translation machinery and allowing increased protein production in the absence of increased transcription. In the cytoplasm, eIF4E recruits mRNAs to the ribosome for initiation of translation. Importantly, eIF4E differentially affects the translation efficiency of mRNAs, where mRNAs with highly structured 5′ untranslated regions (UTRs) are regulated by eIF4E, and housekeeping transcripts such as GAPDH are not. eIF4E is found in much lower amounts than other translation factors in normal cells, so that its availability, negatively regulated by 4E binding proteins (4EBPs), is a rate-limiting factor in initiating the translation of target genes. Phosphorylation of 4EBP1 following mitogenic stimuli, via the mTOR pathway, releases eIF4E, making it available for binding to eIF4G and allowing assembly of the eIF4F translation factor (eIF4E  +  eIF4G  +  eIF4A) (Fig. 14.6).

Overexpression of eIF4E is associated with oncogenic transformation in cell culture and with tumor formation, metastatic disease, and increased tumor invasion in mice. Even moderate overexpression of eIF4E promotes cellular proliferation, growth, and malignant transformation. This can be explained by the fact that the specific mRNAs targeted by eIF4E include many that are critical to cell division, cell growth, and angiogenesis, such as cyclin D1, survivin, c-myc, VEGF, Mcl-1, and bFGF-2 [58]. Overexpression of eIF4E occurs in AML and this overexpression impedes differentiation of leukemia cells. It can be speculated that eIF4E-overexpressing leukemic cells have become dependent on eIF4E for growth and survival and thereby have an addiction to eIF4E.

Several approaches to target eIF4E therapeutically have been proposed. Ribavirin is a well-characterized broad spectrum nucleoside analog with antiviral activity against a range of RNA viruses. Ribavirin targets eIF4E directly in a variety of systems, thereby inhibiting translation and/or mRNA export of transcripts that contain long and highly structured UTRs. A phase I/II clinical trial examining the efficacy of Ribavirin treatment in poor prognosis M4 and M5 AML patients demonstrated clinical activity and associated molecular responses. This provided proof of principle that targeting eIF4E leads to dramatic clinical improvements in this poor prognosis patient population [59].

In conclusion, it is becoming apparent that aberrant control of mRNA translation is an important force underlying tumorigenesis. This has important therapeutic implications, since the molecular network regulating mRNA translation is accessible for pharmacological manipulation.

Acute promyelocytic leukemia (APL) is a subtype of acute myelogenous leukemia, representing 5–8 % of AML cases in adults. At the genetic level, APL is characterized by a specific chromosomal rearrangement between the retinoic acid receptor alpha (RARA) on chromosome 17, and a number of partners. The majority of patients (98 %) presents with the 15;17 translocation, t(15;17), which results in a fusion of RARA with the promyelocytic leukemia (PML) gene on chromosome 15 [60, 61]. In normal cells, the retinoic acid receptors (RARs) heterodimerize with the retinoid X receptors (RXRs) to form a transcriptional complex. In the absence of ligand, all-trans retinoic acid (RA), a vitamin A derivative, the heterodimers are found bound, along with inhibitory corepressor molecules, to the retinoic acid response elements (RAREs) of target genes. Treatment with ligand causes a release of the corepressors and a concurrent recruitment of coactivators, leading to the initiation of transcription of genes important for stimulating myeloid differentiation and regulating the cell cycle.

In APL cells, the prevailing view is that the leukemic effects of the resulting chimeric protein, X–RARA, are due to its function as a dominant negative inhibitor of normal retinoid receptor function. The chimera locates to promoters normally regulated by RARA, aberrantly recruits corepressor proteins, and thereby inhibits the RARA-mediated gene expression. Therapeutic doses of RA can circumvent the differentiation block [62].

To date, seven different chromosomal translocation partners have been identified in patients with APL (Fig. 14.8), and, as mentioned previously, all involve the RARA gene on chromosome 17q21. The N-terminal fractions contributed by the fusion partners donate additional dimerization domains. These acquired dimerization domains, together with the DNA-binding domain of RARA, are required for the oncogenic effect of the fusion proteins and promote formation of chimeric receptor homodimers and provide additional corepressor binding domains. The resulting X–RARA fusion proteins have been shown to recruit the HDAC, NCOR1, and NCOR 2 (NCoR) complex, DNMT1, DNMT3A, repressive histone methyltransferases, and polycomb group proteins.

However, disruption of the normal RARA signaling pathway does not solely account for the pathogenesis of APL. As mentioned previously, the differentiation block mediated by the X–RARAs is necessary, but not sufficient to cause APL. This suggests that a second transformative event is required for full neoplastic development. Mutations in oncogenes such as FLT3 can cooperate with the fusions to provide the second hit necessary to generate leukemia. The presence of these second mutations might also explain why APL cannot be cured by differentiation therapy with ATRA alone, but is highly curable by combinations of ATRA and cytotoxic chemotherapy. Additionally, some translocations generate the reciprocal RARA–X fusion genes which can result in the coexpression of their transcripts in leukemic blasts [63–65]. Important roles in oncogenesis are now being identified for the coexpressed RARA–X reciprocal fusion proteins [63, 64, 66, 67]. Additionally, the partner proteins in APL have important growth-regulatory roles in normal myeloid cells. It may be that the loss of one allele of the partner protein combined with the effects of the X–RARA or reciprocal RARA–X fusion protein, which compromises normal partner protein function, contribute to the development of leukemia.

Core-binding factor (CBF) leukemias refer to a subset of AML bearing t(8;21)(q22;q22) and inv(16)(p13q22). These genomic rearrangements are characterized by disruption of the RUNX1 gene at 21q22 and the CBFbeta (CBFB) gene at 16q22, respectively. These are two of the most prevalent cytogenetic subtypes of AML. Native RUNX1 and CBFB form a heterodimeric transcription factor [68] required for the development of definitive hematopoiesis, including macrophage colony-stimulating factor (CSF) receptor [69], granulocyte–macrophage CSF [70], myeloperoxidase [71], and interleukin-3 [72]. However, the normal function of the ­heterodimer is abrogated in the CBF leukemias.

The MLL (mixed lineage leukemia) gene is located on chromosome 11q23, and fusions involving this gene are seen in both de novo and therapy-related acute myeloid and lymphoid leukemia. MLL is a large DNA-binding protein ubiquitously expressed in hematopoietic cells, including stem and progenitor populations. Using chromatin immunoprecipitation (ChIP) analysis, MLL has been found to be associated with a subset of transcriptionally active human promoters [93, 94] and with RNA polymerase II, suggesting that MLL has a specific role in the regulation of transcription. A SET domain is located at the carboxy-terminal of MLL, and this domain mediates methylation of histone H3 lysine 4 (H3K4), a histone modification which is associated with transcription at active gene loci [95]. Importantly, in mammals, MLL positively regulates the expression of the homeobox (HOX) genes. HOX genes are transcription factors that participate in the development of multiple tissues, including the hematopoietic system.

Some of the roughly 50 characterized MLL fusion partners (FPs) can be grouped into families based on cellular localization and function (Table 14.3) [96, 97]. All identified MLL fusions contain the first 8–13 exons of MLL and a variable number of exons from the FP gene. Furthermore, another type of MLL rearrangement, MLL–PTD (partial tandem duplication), is a result of internal tandem duplication of select exons. MLL mutant proteins are always in-frame chimeras that reside in the nucleus, regardless of whether the fusion partner is normally nuclear or cytoplasmic in origin. Despite the large number and functional diversity of the fusion partners, there are some common principles that can be applied to all MLL fusions [96]. First, all MLL fusions retain its amino-terminal domains required for the association of MLL with chromatin, so the fusion proteins are still able to bind DNA. Second, expression of the amino-terminal region of MLL (the region retained in MLL fusions) alone does not lead to myeloid transformation [98], indicating that the fusion partners make critical contributions to the oncogenicity of MLL fusions [99]. Third, MLL fusion proteins enforce continuous expression of the HOX genes, HOXA9 and MEIS1, which are normally downregulated during hematopoietic differentiation. This sustained induction appears to be critical for leukemogenesis, since forced expression of HOXA9 and MEIS1, in the absence of an MLL fusion, immortalizes hematopoietic progenitors in vitro and results in AML development in transplanted mice [100, 101].

MLL fusions are hypothesized to disrupt normal gene expression patterns, especially those of the HOX genes, maintained by wild-type MLL. However, discerning one unifying mechanism of leukemogenesis by MLL fusion proteins is difficult due to the heterogeneity of the MLL fusion partners. The putative mechanisms of MLL fusion-induced transformation may be more easily characterized by dividing the fusion partners into two broad classes based on their normal cellular localization (either nuclear or cytoplasmic). The fusion partners that are nuclear have been associated with various aspects of transcriptional regulation. Therefore, the fusions involving these proteins may lead to transcriptional deregulation of the HOX genes through alterations in the histone modification pattern and chromatin structure (Fig. 14.9). An early insight into this possibility was provided by the characterization of the nuclear MLL–CREBBP fusion [102]. CREBBP (also known as CBP) is a well-characterized global transcriptional activator with intrinsic histone acetyltransferase (HAT) activity. Structure–function analysis demonstrated that inclusion of the portion of CREBBP that contains its HAT domain in the fusion is required for full in vitro transformation and is sufficient to induce the leukemic phenotype in vivo [103]. These data suggest that the leukemic effect of MLL–CREBBP results from the combination of the chromatin association and modifying activities of CREBBP with the DNA-binding activities of MLL. Additionally, although the SET domain, which mediates MLL’s H3K4 methyltransferase activity, is consistently lost in all of the MLL fusions, several MLL fusion partners, for example, AF4, AF9, AF10 and ENL [104], associate with the DOT1L histone methyltransferase that methylates lysine 79 residues in histone H3 (H3K79). Methylation of H3K79 is also associated with positive transcriptional regulation [105]. As different methylation marks may positively control transcription in unique ways [106], the replacement of H3K4 activity in wild-type MLL with H3K79 activity in the MLL fusion complex could perturb transcriptional control. In the case of the MLL–PTD, where the SET domain is maintained, the characterization of a mouse model of Mll–PTD demonstrated that MLL–PTD facilitates histone H3K4 trimethylation as well as H3/H4 acetylation within target Hox gene promoters [107]. This provides further evidence for an epigenetic mechanism as the underlying cause of Hox overexpression.