Pediatric Myeloid Leukemia, Myelodysplasia, and Myeloproliferative Disease

Clonal Origin of Myeloid Leukemia Cells

Normal myelopoiesis is a complex differentiation program whereby primitive hematopoietic stem cells (HSCs) develop along a multistep pathway into fully differentiated, functionally active circulating blood cells. This exquisitely controlled process is regulated by the intricate interactions among the expression levels of various transcription factors, growth factors and their receptors, cytokines, enzymes, and still unidentified novel molecules. A series of sequential genetic abnormalities perturbs this normal developmental progression and leads to AML. Although the relationships between specific morphologic subtypes of AML and their specific recurring genetic abnormalities have provided some insight into the mechanisms of leukemogenesis, understanding of the process whereby these fusion gene products interact with normal signal transduction pathways to subvert hematopoietic cell development is incomplete.

Several strong lines of evidence support the hypothesis that AML progresses from a single transformed hematopoietic stem or progenitor cell. More than 30 years ago, studies were pioneered by Beutler and colleagues and Fialkow using X inactivation patterns in female patients to establish the clonal origin of human malignancies, including leukemia. By showing the presence of a single glucose-6-phosphate dehydrogenase (G6PD) isoenzyme in the leukemic myeloblasts of heterozygous females, the clonal origin of myeloid leukemia cells was demonstrated. Females who are heterozygous at the G6PD locus express two isoforms of the enzyme. Approximately half of the cells in normal somatic tissue have randomly inactivated one of the X chromosomes, and approximately half of the cells should therefore express each isoform. Unlike normal somatic cells, AML cells of female patients heterozygous for G6PD expressed only one G6PD isoform, indicating cells of clonal origin. In the 1980s, Vogelstein and coworkers developed a strategy using X chromosome–linked DNA restriction length fragment polymorphisms to determine the clonal origin of human tumors. They established that maturing granulocytic cells arise from the malignant clone in patients with AML. Later, X chromosome inactivation to determine clonality in malignancies was carried out using the polymerase chain reaction (PCR) assay to distinguish X-linked polymorphic genes.

Transformation of Stem Cells with Self-Renewal Capacity

These earlier studies, and those performed by Bonnet and Dick and associates in the 1990s, led to the insight that leukemia cell populations form a hierarchy in which leukemia stem cells or leukemia-initiating cells (LICs) are responsible for their own self-renewal and for the generation of the more differentiated progeny within the leukemic clone. This theory was in contrast to the stochastic model, which stated that malignant properties within cancer cell populations follow Gaussian distributions, resulting in the random or stochastic probability that individual cancer cells within the population are able to initiate malignant progression. Whereas AML was the first human cancer in which cancer stem cells were identified, the case has now been made in breast cancer, brain tumors, and colon cancer, among others. In general, AML stem cells are believed to arise through the transformation of HSCs, which then retain the capacity for self-renewal, or through transformation of more differentiated hematopoietic progenitor cells that acquire the ability to self-renew by virtue of the particular transforming mutations, such as the formation of an MLL fusion gene.

AML is a heterogeneous disease in its clinical manifestations, response to therapy, and molecular genetics; thus insight into the cell of origin would have important ramifications regarding diagnosis and treatment. Moreover, refractory disease and relapse are often attributed to the presence of an LIC population that is resistant to chemotherapy; thus the ability to identify and target this subpopulation may be key to improving overall disease outcome. In the 1990s, Dick's group produced a series of convincing papers that strongly implicated the primitive, pluripotent HSC as the LIC in some types of AML. “Stemness,” or the capacity to function as an LIC, includes self-renewal, proliferation, and differentiation and is tested in vitro by replating limiting dilution studies and in vivo by serial transplantation assays with the ability to regenerate a clonal leukemia in secondary and subsequent recipient animals. The HSC, with its inherent gene programs ensuring self-renewal, proliferation, and survival appears to be an ideal target to become sabotaged through a series of genetic events, leading ultimately to a fully transformed AML LIC that is capable of producing a clonal population of leukemia cells. According to this argument, different AML phenotypes often occur as a result of the gene(s) that are altered, leading to differentiation arrest within the progeny of the LIC, but not necessarily reflecting the degree of the commitment of the initially transformed cell. Support for this hypothesis initially came from several clonality studies in leukemic cells from patients with AML, which demonstrated multiple-lineage involvement in a high proportion of cases, and implicated a multipotent HSC as the cell of origin. More supportive evidence has come from studies looking at cytogenetic markers and characteristic cell surface antigen expression patterns. Pluripotent HSCs express CD34 but do not express CD38 or HLA-DR, whereas more committed myeloid progenitor cells are CD34+, CD38+, and HLA-DR+. Using fluorescence-activated cell sorting (FACS) of leukemic cell populations, the same two subpopulations, CD34+/CD38− and CD34+/CD38+, were isolated from the bone marrow of patients with different AML subtypes. The two subpopulations were then evaluated using fluorescence in situ hybridization (FISH) to determine the presence of cytogenetic abnormalities. The studies detected the same characteristic cytogenetic abnormalities in the CD34+/CD38− stem cell fraction as in the original leukemic bone marrow samples, implying origin in a very early HSC compartment.

Further evidence supporting the stem cell origin model of AML was derived from transplantation experiments in which purified human AML cells were transplanted into mice with severe combined immunodeficiency (SCID). These experiments defined a SCID mouse leukemia–initiating cell (SL-IC) in the bone marrow of patients with AML and showed that these SL-ICs are CD34+/CD38− and that their engraftment produces large numbers of colony-forming progenitors. The CD34+/CD38+ and CD34− fractions did not engraft. Similar results were obtained through transplantation studies in a modified SCID mouse, the nonobese diabetic mouse with SCID (NOD/SCID), with cells from patients with AML. These studies identified and analyzed AML stem cells from samples of different FAB subtypes on the basis of their ability to initiate human AML after transplantation in NOD/SCID mice. These SL-ICs were able to proliferate and differentiate after transplantation, producing disease in the mice identical to that in the donor, as well as being able to renew themselves, reestablishing AML in secondary recipients, demonstrating that SL-ICs are capable of self-renewal. Again, the SL-ICs were found to reside in the CD34+/CD38− fraction and not in the CD34+/CD38+ or CD34− fractions. The SL-IC phenotype was consistent regardless of the FAB subtype (e.g., M1, M2, M4, M5). As few as 2 • 10 ⁴ CD34+ cells were able to initiate the leukemic clone in recipient mice, whereas 100 times as many CD34− cells failed to engraft. Cells with the CD34 surface antigen were further fractionated on the basis of CD38 expression. Only the CD34+/CD38− fraction contained SL-ICs.

Whereas the transformation of the HSC provides an attractive mechanism to generate hierarchical myeloid leukemia cell populations, several studies have given new credence to a second mechanism through which committed progenitor cells along the pathway of myeloid differentiation are vulnerable to transforming mutations that confer self-renewal properties to progenitor cells that inherently lack this ability. These transformation events create an LIC capable of generating a population of aberrant cells blocked at the differentiation state at which the initial transforming event occurred. Several translocations involving the MLL gene have been shown to initiate self-renewal in myeloid progenitor cells and to result in an AML phenotype. A head-to-head comparison was conducted by retrovirally transducing the human AML-associated MLL-ENL fusion gene into murine HSC, common myeloid progenitors (CMPs), and granulocytic-monocytic–restricted progenitors (GMPs) and evaluating the relative transforming ability and resultant phenotype in each cell type. AML developed in mice transplanted with any of the three transduced cell populations and, in each case, displayed an identical myelomonocytic phenotype as assessed by morphology and flow cytometry. Similarly, the MOZ-TIF2 fusion gene was able to confer self-renewal properties to committed myeloid progenitor cells and resulted in AML when transplanted into irradiated mice. Although the fusion genes MLL-ENL and MOZ-TIF2 appear to impart leukemogenic capabilities to committed progenitor cells convincingly, other fusion genes such as BCR-ABL can induce leukemic transformation only when expressed in an HSC but not a more mature myeloid cell. Similarly, the potent oncogenic combination of Hoxa9 and Meis1 only resulted in AML when transduced into HSCs but not in more committed myeloid progenitors. In this case, expression of β-catenin was found to be the necessary factor for oncogenesis. Present in HSCs but absent in more committed progenitors, the addition of β-catenin to progenitors transduced with Hoxa9 and Meis1 was sufficient to lead to AML. These findings point to the differing transforming abilities of different oncogenes that are likely cell-type and cell-context specific. Hence both cell-of-origin hypotheses appear to be correct, depending on the specific oncogene or tumor suppressor that provides the initiating genetic lesion. Increased proliferation, survival, and self-renewal are all necessary attributes of a leukemia stem cell. Some genetic abnormalities, such as MLL-ENL and MOZ-TIF2 , may be capable of inducing self-renewal and thus are able to transform a more differentiated myeloid cell into an LIC. By contrast, other oncogene fusions, such as BCR-ABL , or oncogenic transcription factor genes, such as Hoxa9 , do not possess self-renewal capabilities and thus require a cell that inherently has this characteristic, namely, the HSC.

More recently, challenges to the cancer stem cell hypothesis have arisen out of a number of studies, predominantly in solid tumors such as melanoma, where the frequency of cells capable of giving rise to a new tumor in an immunocompromised host is of sufficient frequency to call into question whether the LIC represents a unique subpopulation or the original stochastic model of cancer progression holds true. These studies have again called into question previously accepted tenets. CD34+/CD38+ populations, in addition to the CD34+/CD38−, have also been found to contain LICs when injected into more immune-deficient hosts that lack key components of the innate immune system. These studies and additional syngeneic and cogenic transplant studies in mice as well as zebrafish highlight the critical context of the recipient animal in determining the LIC. The heterogeneity of LICs in AML is further demonstrated by an increasing number of cell surface markers in addition to CD34 and CD38, including CD123, TIM3, and CD47. Additionally, a number of studies have attempted to identify a specific LIC genetic signature, which has been found at least in some cases to independently predict prognosis. In this case, the LIC signature was reminiscent of the normal HSC, hearkening back to the original predictions of Dick's group highlighting the HSC as a putative LIC cell of origin. However, although these findings are provocative, a consistent cadre of LIC genes has not yet been established. Clonality studies demonstrating the progressive acquisition of mutations in a given cell have also been put forward as an alternative to the cancer stem cell hypothesis. Although seemingly at odds, these different concepts can be reconciled into a single integrated framework. Myeloid (and lymphoid) leukemia-initiating events may occur in both HSCs and committed progenitor cells, and each leukemia arises from a combination of multiple genetic abnormalities. Secondary events subsequently are acquired in different subpopulations, generating unique clones with varying malignant potential depending on the genetic lesion and the self-renewal capacity of the cell in which these abnormalities occur. The number of genetic lesions necessary to cause frank AML is being defined based on next-generation sequencing studies recently published by the group at Washington University in St. Louis as well as other centers. The evolution of particular clones appears to determine the eventual phenotype of the AML cells documented at diagnosis ( Fig. 51-1 ).

These new discoveries have added additional complexity to our understanding of the contribution of LICs to AML pathogenesis by demonstrating that the heterogeneity inherent in this disease extends to the AML stem cells. However, the identification of new LIC surface markers, genetic signatures, and gene-specific mutations hold out promise that LIC-targeted therapies will be realized in the coming years.

Class I Mutations

Among the earliest examples of class I mutations, RAS mutations were described in the early 1980s. There are three functional RAS gene family members— NRAS, KRAS, and HRAS —that all act as guanosine nucleotide phosphate (GTP)-binding proteins. Mutations of NRAS are the most prominent RAS mutations in AML and have been found in as many as 30% of adult patients with AML. Mutations, most frequently in codons 12, 13, and 61, result in the retention of the RAS protein in its active GTP-bound state and lead to the constitutive activation of downstream effector proteins, which causes the transcriptional activation of various target genes that direct cellular differentiation, proliferation, and survival ( Fig. 51-3 ). RAS proteins may be alternatively activated by mutations in genes encoding a number of regulatory factors. Activating point mutations in the SHP-2/PTPN11 phosphatase provide a stimulatory signal through guanine nucleotide exchange factors, such as “Son of Sevenless” (SOS), resulting in increased RAS pathway signaling, as do inactivating mutations of neurofibromin-1 (NF1) , which codes for a GTPase-activating protein (GAP) of the same name. NF1, normally functions, at least in part, by increasing the GTPase activity of RAS, thereby inactivating RAS-GTP by converting it into the inactive RAS-GDP form. These three types of mutations function independently to activate RAS signaling and lead to disturbances in hematopoietic cell differentiation and proliferation that may ultimately contribute to the development of MDS, myeloproliferative disease (MPD), and AML (see Fig. 51-3 ). Mutations in NF1 and PTPN11 have been particularly associated with JMML (see later).

There has been significant disagreement regarding the prognostic implications of RAS mutations in MDS and AML. Several studies have demonstrated that RAS mutations in patients with MDS confer a poor prognosis and increase the risk for progression to acute leukemia. Other studies have reported no difference in survival in AML patients who carried RAS mutations compared with patients without RAS mutations, whereas some have suggested improved survival in patients with AML and NRAS mutations. A large study of more than 2500 patients with AML demonstrated no difference in prognosis in patients with or without RAS mutations. This study also demonstrated an association between RAS mutations and a leukemia karyotype containing inv(16), a class II mutation resulting in the CBFβ-MYH11 fusion protein and altered function of the core-binding factor (CBF) transcriptional complex. A recent report from the COG examining 825 pediatric AML samples treated on two recent trials identified NRAS mutations in 10% of patients predominantly in codons 12 and 13, with no mutations in codon 61. Interestingly, most of these mutations were associated with good prognosis NPM1 mutations (see later) and not CBF alterations. However, multivariate analysis confirmed a lack of independent impact of NRAS mutations on prognosis.

Another group of class I mutations involves constitutive activation of receptor or cytoplasmic-nuclear tyrosine kinases. The BCR-ABL translocation that is the sine qua non of CML is an example of a class I mutation causing the activation of the ABL tyrosine kinase. Similarly, receptor tyrosine kinase mutations activate the human protooncogenes KIT and FLT3 in AML. The KIT protein is activated by the binding of its ligand stem cell factor and is critical for the development and growth of mast cells, melanocytes, HSCs, and the interstitial cells of Cajal. Mutations in KIT allow ligand-independent activation of KIT and confer factor-free growth and tumorigenicity in hematopoietic cell lines. Blasts from patients with AML express KIT on their cell surfaces in most cases. Deletional and insertional mutations of KIT have been identified in the blast cells of patients with AML, often in association with the CBF abnormalities inv(16) or t(8;21).

The FLT3 gene encodes a class III fms-like receptor tyrosine kinase expressed in early hematopoietic progenitors. FLT3 mutations occur in two varieties, internal tandem duplications (ITDs) in the juxtamembrane region and activation loop mutations (ALMs), primarily at codons 835 and 836. The FLT3 -ITD is one of the most frequent abnormalities in adult AML, documented in 25% to 30% of patients, and generally confers a poor prognosis. By contrast, ALMs occur in only 10% of adult patients and have not been associated with inferior survival rates. FLT3 -ITDs have been observed in 5% to 16% of pediatric AML cases and ALMs in 2% to 5% of cases. Paralleling the adult studies, FLT3 -ITDs, but not ALMs, were associated with a poor prognosis in children. FLT3 -ITDs result in the constitutive activation of FLT3 and cause interleukin-3–independent growth in Ba/F3 and 32D cells. This activation, in turn, results in the signaling of multiple downstream pathways, including the aforementioned RAS pathway, and the inhibition of caspase-mediated apoptosis through the stimulation of phosphatidylinositol 3′-kinase (PI3K) and AKT (see Fig. 51-3 ). However, mice transplanted with FLT3-transformed hematopoietic cells develop a myeloproliferative disorder characterized by splenomegaly and leukocytosis but they do not develop AML. FLT3 -ITDs often occur in conjunction with other mutations, including NPM1 (see later) and WT1 (see later).

Class II Mutations

In class II mutations, the molecular defect appears to be at the level of transcriptional activation, whereby transcriptional repressors cause a dominant negative inhibition of normal hematopoietic cell differentiation. AML can be subtyped based on the type of class II mutations in the blast cells of patients. The type 1 subset of class II mutations occurs in patients with de novo AML. These are typical chromosomal translocations resulting in chimeric oncoproteins that in some cases may cause the inhibition of differentiation. The type 2 subset of class II mutations usually are found in patients with AML arising after a prodrome of MDS and manifest more commonly in older adults or in patients who have received previous chemotherapy. Cytogenetic studies of the blast cell populations in these patients have complex karyotypes, lacking translocations but harboring deletions such as 5q−, monosomy 7, and 20q−.

Type 1 Mutations

Translocations Involving Core-Binding Factor Complex.

Several of the translocations that have been identified in adult and childhood de novo AML involve the CBF complex. The CBF complex is the most frequent target of chromosomal translocations in the human leukemias. The CBF regulatory complex consists of a DNA-binding subunit, RUNX1 (also called CBFα or AML1) and CBFβ, a subunit that does not bind DNA independently but heterodimerizes with RUNX1 or one of its closely related family members. Chromosomal translocations that modify the CBF complex in AML include the following: t(8;21)(q22;q22), which generates the AML1-ETO fusion protein, more recently referred to as RUNX1-RUNX1T1 ( Fig. 51-4 ); t(3;21)(q26;q22), which gives rise to AML1-EVI1; and inv(16)(p13;q22), which fuses the CBFβ to smooth muscle myosin heavy chain (SMMHC), resulting in the CBFB-MYH11 fusion gene. The AML1 gene was inactivated in the germline of mice and was shown to be essential for definitive hematopoiesis of all cell lineages. Homozygous animals die early in embryogenesis of central nervous system (CNS) hemorrhage, but they have normal morphogenesis and yolk sac hematopoiesis lineages. Inactivation of the CBFβ gene in the mouse has shown a similar phenotype in the homozygous null mice. These experiments have demonstrated that the AML1/CBFα complex is essential for normal hematopoiesis and that chromosomal rearrangements involving this complex may interfere with its regulatory function in ways that lead to disruption of cellular differentiation, and eventually malignant transformation.

Evidence primarily from knock-in studies in mice has supported the hypothesis that the AML1-ETO fusion protein functions as a dominant negative inhibitor of wild-type functions. In these experiments, the AML1-ETO fusion gene is “knocked in” to the wild-type AML1 locus. The phenotype of these mice is identical to the AML1 knockout; the mice die early in embryogenesis from CNS hemorrhage and lack any evidence of definitive hematopoiesis, although the rare cells that survive and express AML1-ETO have an increased capacity for self-renewal. Biochemical experiments have shown that the ETO portion of the AML1-ETO chimeric protein recruits nuclear corepressor complexes to the CBF promoters, resulting in transcriptional inhibition of target genes normally activated by the AML1/CBFβ heterodimeric complex (see Fig. 51-4D ). Subsequent studies using inducible murine model systems to bypass the embryonic lethality of AML1-ETO overexpression have demonstrated that overexpression of this transgene at later developmental time points confers enhanced self-renewal capacity to bone marrow progenitors, as demonstrated by serial culture experiments. Importantly, the mice do not develop AML but appear to be predisposed to leukemogenesis. Thus after treatment with N -ethyl- N -nitrosourea (ENU), the mice developed rapid onset of AML compared with similarly exposed wild-type mice. Moreover, as noted, there is an association in myeloid leukemia of AML1-ETO translocations and mutant tyrosine kinases, such as KIT. Interestingly, AML1-ETO transcripts may persist for a long time in patients with AML1-ETO –expressing AML in sustained remission.

Translocations Involving the Retinoic Acid Receptor-α.

Reciprocal chromosomal translocations involving the retinoic acid receptor-α (RARA) gene locus on chromosome 17 are the defining molecular features of acute promyelocytic leukemia (APL). Before the RARA translocations were described, retinoic acid was known to induce myeloid differentiation in vitro and therapy with all- trans -retinoic acid (ATRA) in patients with APL was shown to induce complete remission (CR). The intense investigation of the RARA gene triggered by these observations culminated in identification of the promyelocytic leukemia PML-RARA translocation breakpoint, t(15;17)(q22;q11), reported simultaneously by several groups. Although most APL cases have the associated t(15;17), other reciprocal translocations involving the RARA gene have been identified in which the gene is fused to other gene partners. The most common of these include the promyelocytic leukemia zinc finger (PLZF) gene on chromosome 11 in t(11;17)(p13;q11) and the nucleophosmin gene (NPM) on chromosome 5 in t(5;17)(q31;q11).

Normally RARA functions as a ligand-dependent activator of transcription. It forms a heterodimer with the related protein, retinoid X receptor (RXR), and binds through its zinc finger domain to a specific promoter sequence; in the presence of the retinoic acid ligand, the RARA-RXR heterodimer activates transcription of retinoic acid–responsive genes. This process is accomplished in concert with a coactivator complex that includes proteins such as p300 and pCAF. In the absence of retinoic acid, a corepressor complex is recruited, composed of N-CoR (nuclear receptor corepressor) or SMRT (silencing mediator of retinoid and thyroid receptors), which binds to another protein called Sin3 and then to HDAC1. HDAC1 is a histone deacetylase protein, which epigenetically alters histones to keep DNA in an untranscribable form. The expression of retinoic acid–responsive genes is essential for normal myeloid development.

Much evidence supports the hypothesis that the PML-RARA fusion protein functions as a dominant negative inhibitor of the PML protein and RXR. PML proteins are normally located in macromolecular nuclear organelles, called PML oncogenic domains (PODs). The PML-RARA fusion protein disrupts the PODs, causing normal PML, RXR, and other nuclear proteins to disperse in an abnormal pattern. ATRA and arsenic trioxide (As ₂ O ₃ ) have shown activity in APL blast cells and in patients with APL, and studies of these drugs have afforded important insights into the mechanisms of leukemogenesis in APL. ATRA and As ₂ O ₃ appear to act through different biochemical pathways, and promyelocytes resistant to ATRA are often sensitive to treatment with As ₂ O ₃ . Both drugs induce degradation of the PML-RARA fusion protein but through different mechanisms. Retinoic acid binding to wild-type retinoic acid receptors (RARs) in the cell nucleus causes degradation of the PML-RARA protein through the ubiquitin-proteosome and caspase systems, thus allowing for the terminal differentiation of leukemic promyelocytes. This binding also results in derepression whereby N-CoR is dissociated from the PML-RARA fusion protein and, subsequently, the nuclear coactivator complex is recruited to reverse histone deacetylase-mediated repression. PODs are relocated back to the normal nuclear pattern. By contrast, As ₂ O ₃ appears to target the PML portion of the PML-RARA fusion protein preferentially and causes its degradation by inducing apoptosis, potentially through downregulation of the antiapoptotic factor BCL2. The activity of these agents has led to the inferences that fusion proteins may interfere with normal myeloid cell development in several ways—through inhibitory effects on assembly of the PODs that contain PML or through dominant inhibitory effects on transcriptional targets of dimeric complexes with normal retinoid receptors, leading to arrest of differentiation at the promyelocyte stage.

Studies using transgenic mice have shown that the PML-RARA fusion product is involved in the development of leukemia. Transgenic mice generated with the PML-RARA fusion protein specifically expressed in the myeloid-promyelocytic lineage develop a myelodysplastic-like disorder during the first year of life. A subset of these mice then develops a form of acute leukemia that closely mimics human APL and responds to ATRA. Given the relatively long latency period and the development of leukemia in only a fraction of the mice, it is likely that the PML-RARA translocation is insufficient by itself to cause APL and that second mutations are necessary for leukemic transformation.

Mixed-Lineage Leukemia Gene Translocations.

Transcriptional coactivators and corepressors have been implicated in leukemogenesis. An example is the myeloid-lymphoid or MLL protein, a large protein containing transcriptional activation and repression domains, which is believed to act epigenetically to modulate gene expression through its effects on chromatin structure and configuration. Wild-type MLL is required for normal hematopoiesis and HSC development. Mice deficient in MLL die on embryonic day 10.5 and have numerous skeletal, neural, and hematologic deficits. Heterozygous mice have defects in embryonic segmentation and yolk sac hematopoiesis, and adults demonstrate a mild anemia and thrombocytopenia. Chromosomal translocations of the MLL gene on chromosome segment 11q23 have been identified in human AML and ALL and confer a poor prognosis. Many gene partners have been identified that are involved in these translocations, including AF4, AF9, ENL, AF6, ELL , and AF10 . In most cases, the fusion proteins incorporate the 5′ end of MLL and the 3′ end of the partner. Despite the number of different fusion partners, leukemias with MLL translocations tend to possess a distinct genetic signature compared with AML and ALL samples lacking an MLL translocation. This genetic signature includes the upregulation of a number of specific HOX family genes, including HOXA9, HOXA7, and MEIS1 . Wild-type MLL is known to play a major role in HOX gene transcriptional regulation. In turn, HOX genes play critical roles in embryonic development and normal hematopoiesis. A number of studies have clearly demonstrated that overexpression of specific major HOX genes in mouse models and human bone marrow samples, whether by involvement in reciprocal translocations or by gene upregulation, is linked to the transformation of malignant HSCs in AML. The mechanism whereby MLL fusion proteins upregulate HOX gene expression may differ, depending on whether the fusion partner encodes a nuclear partner with direct transcriptional activity or a cytoplasmic protein that results in MLL protein dimerization and subsequent transcriptional activation.

Transcriptional control may be mediated through the modification of histone proteins, which function to maintain chromatin structure. Recently, the DOTL1 complex was identified as a critical factor in MLL-induced leukemogenesis. DOT1L is a histone-3-lysine-79 (H3K79) methyltransferase, which was previously believed to play a fairly ubiquitous role in methylating gene targets in concert with transcription. However, Zhang and Armstrong's group definitively demonstrated through a series of elegant in vitro and in vivo experiments a specificity for DOT1L in the epigenetic regulation of MLL-translocated downstream targets. When DOT1L was knocked out or absent from MLL-AF9 –transformed cell lines or transgenic mice, leukemia was prevented. The MLL gene is rearranged in up to 20% of pediatric cases of AML. Translocations of MLL are the single most common genetic alteration in infants with acute leukemia, regardless of phenotype, and account for approximately 70% of all cases of AML and ALL in infants. In pediatric and adult cases of AML, the presence of the 11q23 translocation is generally associated with an unfavorable prognosis.

Although less common than MLL translocations, the MLL gene can be altered by partial tandem duplications (PTDs) of particular exons. These PTDs occur in adult and pediatric AMLs with a normal karyotype, as well as those with trisomy 11. Similar to MLL fusions, the MLL -PTD tends to be associated with a poor prognosis and often with the presence of an FLT3 mutation. The mechanisms underlying MLL-PTD leukemogenesis have not yet been elucidated but appear to be different than those of MLL fusions. Interestingly, MLL -PTD appears to repress wild-type MLL expression, which may be an important feature underlying the leukemogenic potential of this abnormality.

French-American-British Classification

In 1976, the FAB group proposed a system of classification of AML based on morphologic and cytochemical features (see Table 51-1 ). The system divides AML into seven subtypes, M1 through M7. An M0 subtype was later added to describe undifferentiated leukemia and, since 1976, immunophenotypic data have also been included. In general, AML can be differentiated from ALL based on morphologic features and cytochemical stains. Cells of myeloid origin should stain with myeloperoxidase and SBB. Cells of monocytic derivation usually stain with NSE. AML blasts usually do not have PAS activity, except for erythroblasts in AML-M6 and eosinophils of the M4Eo subtype.

World Health Organization Classification

The discovery of cytogenetic and molecular genetic abnormalities in malignant disease has had a significant impact on our understanding of malignant transformation. In myeloid malignancies, some genetic abnormalities have prognostic significance whereas others appear to define specific disease subtypes. The WHO has proposed a classification system for neoplastic diseases of the hematopoietic and lymphoid tissues that includes a classification for AML (see Box 51-2 and Fig. 51-5 ). This classification uses the traditional FAB-type morphologic categories of disease and includes additional entities such as immunophenotype, molecular genetic, and clinical characteristics that make the system more clinically relevant for diagnosis, prognosis, and treatment. The WHO classification system divides myeloid diseases into several major subtypes—myeloproliferative neoplasms; myeloid neoplasm associated with eosinophilia and abnormalities of PDGFRA , PDGFRB , or FGFR1 ; myelodysplastic-myeloproliferative neoplasms; MDS, AML and related neoplasms; and acute leukemias of ambiguous lineage. The 2008 revision of the WHO classification has expanded the category of AML to include AML with recurrent cytogenetic translocations, AML with myelodysplasia-related changes, therapy-related myeloid neoplasms, AML not otherwise specified, myeloid sarcoma, myeloid proliferations related to Down syndrome, and blastic plasmacytoid dendritic cell neoplasms.

Within the subtype containing recurrent cytogenetic translocations, seven additional categories have been described:

• AML with t(8;21)(q22;q22); AML1-ETO ; RUNX1-RUNX1T1

• AML with inv(16)(p13.1q22) or t(16;16)(p13.1;q22); CBFB-MYH11

• APL with t(15;17)(q22;q12); PML-RARA

• AML with t(9;11)(p22;q23); MLLT3-MLL

• AML with t(6;9)(p23;q34); DEK-NUP214

• AML with inv(3)(q21q26.2) or t(3;3)(q21;q26.2); RPN1-EVI1

• AML (megakaryoblastic) with t(1;22)(p13;q13); RBM15-MKL1

AML with t(6;9)(p23;q34) and AML with inv(3)(q21q26.2) is relatively uncommon in children. Myeloid sarcomas, formerly called chloromas, may precede hematopoietic evidence of disease and so are now classified as a separate entity. The myeloid diseases of Down syndrome were incorporated into the revised WHO classification in 2008 and are described later in this chapter. Plasmacytoid dendritic cell neoplasms were formerly referred to as blastic NK-cell lymphomas, but the nomenclature has changed to reflect the origin of this myeloid malignancy in this subset of dendritic cells. This rare disease can manifest in children with skin lesions, adenopathy, and ultimately bone marrow involvement. Blast cells characteristically express CD4, CD43, CD56, and CD123 without expression of CD34 or CD117.

The FAB standard used to define AML had been 30% replacement of the bone marrow by the blast cells, but later studies have shown that patients with 20% to 30% blasts (previously classified as refractory anemia with excess blasts in transformation [RAEB-T]) have a prognosis similar to that of patients with more than 30% blasts. The blast count for the diagnosis of AML in the WHO classification was therefore changed to 20%, and the category of RAEB-T was eliminated.

Certain specific and recurrent cytogenetic abnormalities are common in MDS, in alkylating agent–related AML, and in de novo AML with a poor prognosis. These abnormalities include 3q−, −5, 5q−, −7, 7q−, +8, +9, 11q−, 12p−, −8, −19, 20q−, +21, t(1;7), t(2;11), and complex karyotypes. In the classification scheme, these cytogenetic abnormalities were considered to indicate a poor prognosis. Prior therapy with topoisomerase II inhibitors (i.e., epipodophyllotoxins and doxorubicin) was also associated with a poor prognosis. Typically, patients in whom AML develops after exposure to these agents have translocations involving 11q23 ( MLL ). The WHO classification includes these patients in the poor-prognosis category but distinguishes them from having alkylating agent–related secondary leukemia.

Patients may have MDS with unilineage or multilineage dysplasia, and these are reflected as distinct diseases in the most recent iteration of the WHO classification. Dysplasia is defined as being observed in greater than or equal to 10% of the cells of a given myeloid lineage. Children with 2% to 19% peripheral blasts or 5% to 19% bone marrow blasts can be classified according to the same criteria as adult MDS, but a new category of refractory cytopenia of childhood has been introduced to define children with persistent cytopenias and dysplasia of two or more lineages but less than 2% peripheral blasts or less than 5% bone marrow blasts.

Remission Induction

During the past 40 years, survival rates for patients with AML have risen from less than 10% to more than 50% of patients because of intensification of chemotherapy, use of hematopoietic stem cell transplantation (HSCT), and improved supportive care. The first phase of chemotherapy is termed induction and consists of two to four cycles of intensive chemotherapy with the goal of inducing remission. Morphologic remission is defined as the presence of fewer than 5% blasts visible on bone marrow aspirate obtained after recovery of counts. Patients may also achieve cytogenetic and FISH-negative remission, as defined by the absence of a previously detected leukemic clone–associated cytogenetic abnormality in the bone marrow cells. Remission is also now being defined by the absence of MRD, assessed by multiparameter flow cytometry that is designed to detect as few as 0.1% leukemic blasts in the marrow. The inclusion of MRD evaluation has not only identified children at risk for relapse who appeared to be in morphologic remission but has also revealed false-positive recovering marrows that had been deemed refractory disease.

Induction chemotherapy should be initiated as soon as the diagnosis of AML is confirmed, preferably by morphologic, immunophenotypic, and cytogenetic studies of the bone marrow. In some cases, particularly in patients presenting with hyperleukocytosis, the patient's condition may be too unstable for bone marrow evaluation, in which case the diagnosis can often be confirmed on studies of the peripheral blood. Induction chemotherapy regimens consist of high-dose myelosuppressive cytotoxic agents that result in prolonged pancytopenia. The period of bone marrow hypoplasia generally lasts from 21 to 30 days from the initiation of therapy but may last considerably longer. During this period, patients have a high risk for life-threatening infection and bleeding. Aggressive supportive care is necessary (see later).

Postremission Therapy

Several randomized studies have demonstrated that without postremission therapy almost all patients with newly diagnosed AML suffer a relapse within 2 years. The strategies for postremission therapy include autologous HSCT, allogeneic HSCT, or continued courses of chemotherapy. The CCG trials have shown no benefit to autologous HSCT over continued cycles of chemotherapy so, although that strategy continues to be used in some European trials, autologous HSCT is not included in current U.S. trials. Most studies have shown a lower relapse rate for AML patients treated with matched related donor HSCT versus chemotherapy, but the OS is often no better because of the significant morbidity and mortality of HSCT. For this reason, until the most recent trials, unrelated donor HSCT was believed to be too morbid for patients in first remission and thus trials used a genetic randomization, wherein all patients with a matched related donor proceeded to HSCT and all others proceeded with further cycles of chemotherapy.

Based on the success of the MRC trials with chemotherapy alone using a combination of cytogenetic characteristics and response to therapy, the COG and SJCRH adopted risk-based criteria for HSCT. HSCT, even from a matched related donor, is not recommended for patients with low-risk disease. This was previously defined by COG as the presence of t(8;21) or inv(16) in the leukemic clone at diagnosis, but more recently the presence of a CEBPA or NPM1 mutation has been added to this definition, based on the data revealing the better outcome for these patients. The COG has classified high-risk disease as monosomy 7, −5/5q−, FLT3 -ITD with high allelic ratio, MRD positive after induction chemotherapy, or more than 15% blasts in the marrow after one course of induction chemotherapy. * The incorporation of MRD after induction has enabled the COG to relegate normal karyotype patients to high- and low-risk groups, thereby dichotomizing patients and eliminating the intermediate-risk category. Therefore on the current AAML1031 protocol, high-risk patients will receive best available donor HSCT for consolidation, whereas good-risk patients will not go to transplant even if a sibling donor match is available. With recent advances in supportive care and improved therapy to prevent and treat graft-versus-host disease (GVHD), outcomes have improved for matched related and matched unrelated donor HSCT. Therefore for patients with high-risk disease, both trials recommend HSCT in first CR with an unrelated donor if no matched family donor is identified. The optimal number of cycles of chemotherapy before HSCT to balance minimizing toxicity while minimizing MRD is variable, with the MRC10 trial using four cycles, the current SJCRH trial including two to three cycles, and the COG trial using three cycles before HSCT.

* References .

For patients who do not proceed to HSCT there is controversy as to how many courses of chemotherapy are needed and what chemotherapy should be included. Several trials have shown that cycles of high-dose cytarabine after remission are associated with improved long-term survival. ^† MRC12 randomized patients to receive four versus five total cycles of chemotherapy, with early results suggesting that five cycles are superior; thus that strategy was subsequently used in the current U.S. trials. However, more recent data from the MRC have shown that the fifth cycle added toxicity with no additional therapeutic benefit. These findings are reflected in the current COG trial in which only four cycles of chemotherapy are used for good-risk patients with elimination of the high-dose cytarabine and asparaginase cycle (Capizzi II) that previously comprised the fifth cycle on AAML0531 and AAML03P1. High-risk patients for whom a donor is not identified also only receive four cycles of chemotherapy, but they do receive Capizzi II as their last consolidation cycle. The SJCRH trial has tried to make up for the lack of amsacrine by tailoring the first postinduction cycle based on the patient's AML subtype and cytogenetics. On AML02, patients with t(9;11) or inv(16) AML receive cytarabine and cladribine, patients with M4/M5 AML receive cytarabine and etoposide, and patients with t(8;21) and all others receive high-dose cytarabine and mitoxantrone followed by two courses consisting of cytarabine and asparaginase followed by cytarabine and mitoxantrone. However, based on small subgroup numbers, AML08 uses a single backbone for all patients with the addition of sorafenib for patients with FLT3 -ITD (J. Rubnitz, personal communication, January 2013). CNS leukemia in patients with AML is highly sensitive to chemotherapy, and the addition of cranial irradiation does not decrease CNS relapse compared with systemic and intrathecal chemotherapy. Therefore, cranial irradiation is no longer part of the therapy for patients with AML. However, CNS-positive patients defined on the current COG AML study as having even one blast in the cerebrospinal fluid are treated with additional intrathecal cytarabine until the fluid is clear of blasts and for two additional intrathecal treatments beyond. Furthermore, in pediatrics, the addition of prolonged low-intensity maintenance therapy is associated with lower OS because of more resistant disease at relapse, so maintenance therapy is not used in the current U.S. trials.

† References .

Supportive Therapy at Diagnosis and during Therapy

The diagnosis and treatment of childhood AML require a multidisciplinary approach involving pediatric oncologists, hematopathologists, pediatric surgeons, infectious disease specialists, pediatric intensivists, and pediatric nurses. While studies and procedures are initiated to make the diagnosis, special care must be taken to prevent and treat the life-threatening consequences of leukemia and accompanying pancytopenia. The most serious complications of leukemia at diagnosis are infection and bleeding with pancytopenia, tumor lysis syndrome, and leukostasis. The most common causes of death once remission is achieved are relapse, infection, hemorrhage, and cardiac events. The improvement in OS of patients with AML over the past several decades is attributable as much to better supportive care efforts as to more effective treatment regimens.