Acute Megakaryocytic Leukemia

Abstract

Acute megakaryoblastic leukemia (AMKL) is a rare malignancy affecting megakaryocytes, platelet-producing cells that reside in the bone marrow. Children with Down syndrome (DS) are particularly prone to developing the disease and have a different age of onset, distinct genetic mutations, and better prognosis as compared with individuals without DS who develop the disease. Here, we discuss the contributions of chromosome 21 genes and other genetic mutations to AMKL, the clinical features of the disease, and the differing features of DS- and non-DS-AMKL. Further studies elucidating the role of chromosome 21 genes in this disease may aid our understanding of how they function in other types of leukemia, in which they are frequently mutated or differentially expressed. Although researchers have made many insights into understanding AMKL, much more remains to be learned about its underlying molecular mechanisms.

Acute megakaryoblastic leukemia (AMKL) is a subtype of acute myeloid leukemia (AML) that affects megakaryocytes. It is generally subdivided into three groups based on the characteristics of the person who has the disease: children with Down syndrome (DS), children without DS, and adults (typically without DS). Each of these subcategories has a unique subset of disease-causing or -promoting genetic alterations and have different outcomes. DS-AMKL cells are characterized by the uniform presence of trisomy 21 (T21) and GATA1 mutations, which are accompanied by mutations in chromatin regulators such as cohesin subunits and EZH2 or signaling molecules such as those in the JAK/STAT and RAS pathways (Yoshida et al. 2013). In contrast, leukemia cells in pediatric non-DS-AMKL do not harbor GATA1 mutations, but rather generally have chromosomal translocations that result in expression of disease-causing fusion proteins. The prognosis for this group of children without DS is worse than that of those with DS (Radtke et al. 2009; de Rooij et al. 2017). Adults with AMKL tend to have a much worse prognosis than the pediatric groups with overall survival of less than one year (Tallman et al. 2000; de Rooij et al. 2017). The primary disease causing alterations in adult AMKL is unclear, but there are a number of cases with mutations in cohesin and splicing factor genes as well as TP53 and DNMT3A (de Rooij et al. 2017).

Children with DS have a decreased risk of developing most solid tumors, but an increased risk of acquiring hematologic malignancies including AML (Nižetić and Groet 2012; Maloney et al. 2015). In particular, young children with DS are >100 times more likely to develop AMKL (Lange 2000; Hasle et al. 2016). DS-AMKL occurs following several discrete steps, including acquisition of T21, a GATA1 mutation, and a third genetic event which leads to malignant transformation (Fig. 1). The presence of T21 is sufficient to cause various hematologic abnormalities, such as increased numbers of megakaryocytes in utero (Chou et al. 2008; Tunstall-Pedoe et al. 2008; Malinge et al. 2012). However, T21 alone is not sufficient to promote leukemia. Patients with DS-AMKL have another critical event: mutations in the hematopoietic transcription factor GATA1, which result in production of the shortened GATA1s protein that lacks the first 83 amino acids (Wechsler et al. 2002). Together, T21 and a GATA1 mutation lead to transient myeloproliferative disorder (TMD, also known as transient abnormal hematopoiesis, TAM), a self-limiting preleukemic condition seen at birth to varying degrees in up to 30% of DS neonates (Mundschau et al. 2003; Roberts et al. 2013). Although the majority of TMD cases undergo spontaneous remission, 10%–20% of clinically significant cases progress to DS-AMKL within four years (Khan et al. 2011; Maloney et al. 2015). This progression is associated with the acquisition of additional mutations in one or more genes involved in signaling or epigenetic regulatory pathways that lead to malignant transformation of residual GATA1 mutant clones (Yoshida et al. 2013). It is not entirely clear how these additional mutations cooperate with chromosome 21 genes or GATA1s expression, or how they contribute to DS-AMKL. Although AMKL patients with DS have a favorable prognosis, these children are unusually sensitive to chemotherapy, and gaining a better understanding of the molecular mechanisms behind the disease may lead to better-tolerated targeted therapies.

Figure 1.

Genetic events in DS-AMKL. Up to 30% of neonates with Down syndrome (DS) acquire a GATA1 mutation, which leads to exclusive expression of the short Gata1s isoform lacking the first 83 amino acids. Trisomy 21 (T21) and GATA1 mutations are sufficient to initiate transient myeloproliferative disorder (TMD), a preleukemia, but not acute megakaryoblastic leukemia (AMKL), which is accompanied by secondary gene mutations.

TRISOMY 21

Altered Incidence of Malignancies in DS Individuals

The incidence of DS is between 1 in 319 and 1 in 1000 live births throughout the world (Asim et al. 2015). This disorder is associated with intellectual disabilities, physical characteristics such as slanted eyes, flat nasal bridges, and poor muscle tone, and an increased risk of other disorders such as Alzheimer's disease and congenital heart defects. DS is most often caused by the presence of three copies of the chromosome 21, but in ∼4% of cases it is caused by a Robertsonian translocation, in which the long arm of chromosome 21 is fused to the centromere of another chromosome or to another chromosome 21 (Mutton et al. 1996).

A 1957 mail survey first revealed that DS and leukemia occurred together more frequently than was expected by chance (Krivit and Good 1957). One percent of DS children develop a hematologic malignancy—TMD, AML, AMKL, or B-ALL—giving them a 10- to 20-fold excess risk of leukemia (Lange 2000). Children with DS are >100 times more likely than children without DS to develop AMKL (FAB AML-M7) (Lange 2000). More recent studies have indicated that blood cancers are much more common (standardized incidence ratio [SIR] 5.5, 95% confidence interval [CI] 4.2–7.1), including myeloid leukemias (SIR 11.8, 95% CI 7.11–18.5), in the DS population compared with individuals without DS (Hasle et al. 2016). In fact, leukemia represents 95% of the malignancies in the DS group, compared with 34% in children without DS (Hasle 2001). Surprisingly, individuals with DS have a decreased incidence of most solid tumors with the notable exception of testicular cancer, although the biologic basis for this remains unclear (Hasle et al. 2016).

T21 Leads to Perturbed Hematopoiesis and an Increased Propensity for Leukemia

Although multiple genetic hits are necessary in order for acute leukemia to occur, alterations in hematopoiesis occur in the presence of T21 alone. Various mouse models of DS display elevated levels of immature megakaryocytes and platelets, anemia, and extramedullary hematopoiesis (Kirsammer et al. 2008; Alford et al. 2010; Malinge et al. 2012). Similarly, fetal liver hematopoietic stem and progenitor cells (HSPCs) with T21 produce increased numbers of unusually proliferative megakaryocyte and erythroid progenitors (Chou et al. 2008; Tunstall-Pedoe et al. 2008). Highly proliferative megakaryocytes have been identified in T21 fetuses in the absence of GATA1 mutations, and in one study, 195 of 200 neonates with DS had circulating blasts (De Vita et al. 2008; Rougemont et al. 2010; Roberts et al. 2013). It is likely that these increased numbers of early progenitors could lead to additional opportunities for acquisition of transforming mutations.

There are many factors that may contribute to the link between T21 and the amplified incidence of leukemia. Cells with T21 tend to take on several morphological characteristics associated with tumor cells, including chromosome instability, acquisition of aneuploidy on additional chromosomes, formation of telomere aggregates, and high rates of copy number alterations (Nižetić and Groet 2012). Additionally, cells from individuals with DS have increased levels of reactive oxygen species, higher rates of both genomic and mitochondrial DNA damage, and impaired single-strand break repair and base excision repair pathways (Nižetić and Groet 2012). These phenotypes may help create a permissive environment where leukemic mutations occur and persist.

Chromosome 21 Genes with Possible Roles in AMKL

There are multiple genes on chromosome 21 that are linked to hematopoiesis, megakaryopoiesis, and leukemia, and it is likely that altered dosage of these genes contributes to aberrant blood development. Acquisition of additional copies of chromosome 21 is the most common chromosomal abnormality in acute leukemia, and these trisomic clones often disappear on remission (Cavani et al. 1998; Lange 2000). Furthermore, multiple chromosome 21 genes have been linked to megakaryocyte differentiation, and individuals with monosomy 21 have thrombocytopenia and an absence of megakaryocytes (Huret and Léonard 1997).

A prominent chromosome 21 gene that is known to play a role in both megakaryopoiesis and leukemia is RUNX1 (AML1). This transcription factor is up-regulated before megakaryocyte differentiation and is essential for both hematopoietic stem cell (HSC) maintenance and megakaryocyte maturation (Elagib et al. 2003; Growney et al. 2005; Tijssen and Ghevaert 2013). Furthermore, germline RUNX1 mutations lead to familial platelet disorder, and translocations involving this gene are frequently seen in MDS, AML, and ALL (Bellissimo and Speck 2017). In the context of DS, RUNX1 is expressed at high levels in fetal livers in which it fosters expression of early HSPC genes (De Vita et al. 2010). However, when Runx1 was restored to disomy in a mouse model of DS, there was no effect on the megakaryocyte hyperplasia and fibrosis phenotype, suggesting that other chromosome 21 genes are responsible for the perturbed hematopoiesis of DS (Kirsammer et al. 2008).

Other candidate genes that may play a role in DS-AMKL include ERG and ETS2, transcription factors in the ETS family. ERG is induced during megakaryocyte differentiation and binds the promoters of HSC and megakaryocyte genes (Rainis et al. 2005). Furthermore, when ERG is expressed in murine fetal liver HSPCs, it is sufficient to immortalize cells and lead to leukemia when transplanted into mice (Salek-Ardakani et al. 2009). Overexpression of ERG also synergized with GATA1s to enhance megakaryopoiesis (Salek-Ardakani et al. 2009; Stankiewicz and Crispino 2009, 2013). Furthermore, restoration of Erg to disomy within a mouse model of DS reversed the elevated numbers of HSPCs and megakaryocytes typically seen in this model (Ng et al. 2010) and also caused a decrease in the number of immature megakaryocytes generated from T21 iPSCs (Banno et al. 2016). Likewise, ETS2 is important for megakaryocyte differentiation; its expression is elevated in T21 cells, and forced ETS2 expression is sufficient to up-regulate megakaryocyte genes and induce megakaryocyte differentiation in an erythroleukemic cell line (Ge et al. 2008). These genes may each bias HSPCs in DS individuals to form megakaryocytes.

CHAF1B is a gene on chromosome 21 that delivers H3.1/H4 heterodimers to newly replicating DNA as part of the chromatin assembly factor (CAF1) complex. It was first identified as a possible disease-promoting candidate in leukemia through an shRNA screen for chromosome 21 genes in two DS-AMKL cell lines (Malinge et al. 2012). Knockdown of CHAF1B produced an increased population of cells carrying CD42, a marker of mature megakaryocytes, but a decrease in the degree of polyploidy (Malinge et al. 2012). It is highly expressed in multiple subtypes of blood cancers, including myeloid leukemias, in which its elevated expression is linked to worse survival. Overexpression of CHAF1B supports proliferation and blocks differentiation of HSPCs (Volk et al. 2018). It is still unclear what role this gene may play within DS-AMKL, but it is thought that increased levels of CHAF1B may contribute to the block in differentiation of AMKL blasts.

Numerous miRNAs that may play a role in cancer reside on chromosome 21. For example, mir-125b2 is overexpressed in TMD and DS-AMKL and its overexpression enhanced the proliferation and self-renewal of megakaryocyte progenitors (Klusmann et al. 2010b). Synergy with GATA1s was also observed. mir-155 is overexpressed in B-cell lymphomas and targets p53INP1, a p53 regulator (Nižetić and Groet 2012). Other microRNAs on this chromosome include let-7c, miR-99a, and miR-802 (Alexandrov et al. 2018). It is still unclear whether or how these miRNAs contribute to the myeloid disorders of DS.

Although multiple genes on chromosome 21 play a role in both normal and malignant blood development, it is still unknown which genes are specifically required for DS-AMKL. Many of these genes also cooperate with GATA1 mutations and/or additional mutations, as described below in more detail. It is clear, however, that T21 alone is not sufficient for DS-AMKL: GATA1 mutations are nearly always present, and across all types of hematopoietic malignancies in DS individuals, in only 0.4% of cases is T21 the only chromosomal aberration present (Mitelman et al. 1990). Other genetic events are necessary for progression to acute leukemia.

GATA1s

Role of GATA1 in Normal Blood Development

Up to 30% of newborns with DS and virtually all individuals with TMD and DS-AMKL have a detectable mutation in GATA1. This gene, located on the X chromosome, encodes a lineage-determining transcription factor that contains an amino-terminal transactivation domain and two DNA-binding zinc fingers. GATA1 is expressed in megakaryocyte, erythroid, mast, basophil, and eosinophil lineages (Fujiwara et al. 1996). GATA1 is induced in the megakaryocyte-erythroid progenitor (MEP), where its expression is necessary for differentiation down both lineages (Stachura et al. 2006). Embryonic stem cells lacking Gata1 are able to differentiate into erythroid precursors that express several GATA1 target genes, but the cells subsequently undergo apoptosis (Weiss and Orkin 1995). When the gene was deleted in the germline, the mice died in midgestation from anemia (Fujiwara et al. 1996). In contrast, mice in which Gata1 was deleted selectively in megakaryocytes survived but displayed thrombocytopenia with an accumulation of immature megakaryocytes and a progression to myelofibrosis (Shivdasani et al. 1997; Vannucchi et al. 2002). GATA1-deficient megakaryocytes fail to fully differentiate but rather proliferate excessively both in vitro and in vivo (Kuhl et al. 2005; Muntean and Crispino 2005) A different mouse model, in which levels of GATA1 are reduced to 5% of normal, die beginning at 5 mo of age after manifesting a disease with similarities to human myelodysplastic syndrome/myeloid leukemia (Shimizu et al. 2004). Further evidence for a critical role of GATA1 in erythroid cell and megakaryocyte development comes from human studies. Germline mutations in GATA1 are associated with a spectrum of benign disorders including congenital thrombocytopenia and anemia (Crispino and Weiss 2014).

GATA1 Mutations in the Myeloid Malignancies of DS

GATA1 mutations have been identified in nearly all cases of TMD and DS-AMKL (Wechsler et al. 2002; Mundschau et al. 2003). These mutations generally occur within exon 2 and include nonsense, missense, insertion, and deletion mutations and result in alternative splicing of the exon or cause expression of the gene from an alternative downstream translation start site. These alterations uniformly lead to exclusive production of a short isoform termed GATA1s, which lacks the first 83 amino acids of the wild-type protein (Wechsler et al. 2002; Groet et al. 2003; Hitzler et al. 2003; Rainis et al. 2003; Xu et al. 2003). The GATA1 mutation is an early event in DS-AMKL, occurring in utero as determined by its presence in fetuses and neonates (Mundschau et al. 2003; Taub et al. 2004; Roberts et al. 2013). Surprisingly, up to 30% of neonates with DS have GATA1 mutations, many of which are only detectable by sensitive next-generation sequencing methods; often these infants show no clinical signs of TMD and have a disease that has been named silent TMD (or silent TAM) (Roberts et al. 2013). Notably, GATA1 mutations have only rarely been detected in individuals with AMKL without chromosome 21 aneuploidy, other subtypes of AML, or other hematologic malignancies (Wechsler et al. 2002).

GATA1s-expressing megakaryocytes display abnormally increased levels of proliferation in vitro and in vivo (Kuhl et al. 2005; Li et al. 2005; Muntean and Crispino 2005). The mouse model that only expresses GATA1s develops a transient expansion of megakaryocyte progenitors coupled with impaired fetal erythropoiesis (Li et al. 2005). This altered proliferation and differentiation is likely caused by differential chromatin occupancy of GATA1 and GATA1s (Byrska-Bishop et al. 2015; Chlon et al. 2015; Ling et al. 2019).

GATA1s Cooperation with Genes on Chromosome 21

As discussed, the combination of T21 and a GATA1 mutation is sufficient to cause altered hematopoiesis in both in vitro and in vivo (Yoshida et al. 2013; Maloney et al. 2015). There are multiple ways in which GATA1s could interact and cooperate with T21 genes. For example, fetal liver cells from embryos with T21 display elevated expression of both the long and short isoforms of GATA1, and studies using induced pluripotent stem cells (iPSCs), cells derived from neonates with DS, have either seen increased or variable GATA1s expression, indicating that chromosome 21 genes may play a role in regulating the GATA1 expression (Banno et al. 2016). Additionally, although mouse models of DS do not display a TMD-like disease, when bred with GATA1s mutant mice, compound mutant mice display thrombocytosis, increased fibrosis, and splenomegaly (Malinge et al. 2012).

Full-length GATA1 physically interacts with RUNX1 to activate the megakaryocyte-specific αIIb integrin promoter (Elagib et al. 2003). Of note, one study showed that a mutant GATA1 lacking the first 85 amino acids did not retain the capacity to interact with RUNX1, whereas a different study indicated that Gata1s indeed could bind RUNX1, but the ability to bind and activate the GPIbα gene was diminished (Elagib et al. 2003; Xu et al. 2006). Furthermore, GATA1 and RUNX1 share many transcriptional targets, including GPIIβ and GPIbα, JAK2, and the thrombopoietin receptor c-MPL (Goldfarb 2009).

GATA1 binding sites in megakaryocyte progenitors frequently contain ETS-like motifs, which are bound by ETS-family transcription factors such as ERG and ETS2, indicating that these proteins might cooperatively bind targets to regulate megakaryocyte differentiation (Chlon et al. 2012). Indeed, expression of ERG was sufficient to immortalize Gata1s but not GATA1 wild-type murine fetal liver HSPCs, and ERG and ETS2 overexpression each led to elevated numbers of immature megakaryocytes in a wild-type or a Gata1s context (Salek-Ardakani et al. 2009; Stankiewicz and Crispino 2009). ERG expression alone or in conjunction with Gata1s was also shown to be sufficient to cause leukemia in vivo (Salek-Ardakani et al. 2009).

Finally, DYRK1A is known to have a tumor suppressive role in the context of certain solid tumors, possibly by impairing angiogenesis, but an oncogenic role in DS-AMKL (Nižetić and Groet 2012). It is overexpressed in TMD and DS-AMKL blasts, and knockdown of DYRK1A in DS-AMKL cell lines resulted in alterations in megakaryocyte differentiation (Malinge et al. 2012). Additionally, its expression in murine bone marrow cells was sufficient to produce elevated immature megakaryocytes, and this phenotype was further enhanced in a Gata1s context.

CLINICAL AND BIOLOGICAL FEATURES OF TMD AND DS-AMKL

TMD was first defined in 1964, after several neonates with DS presented with a leukemia-like disease, including splenomegaly and leukemic blasts in the peripheral blood and marrow, which quickly and permanently resolved (Engel et al. 1964). The disease is characterized by increased numbers of small, dysplastic megakaryocytes, and blasts with megakaryocytic features in the peripheral blood and liver. It also may be accompanied by either thrombocytopenia or thrombocythemia, low WBC counts, and, rarely, anemia (Zipursky et al. 1997). Some studies have further identified the presence of elevated counts of immature eosinophils or basophils, and studies have found that TMD blasts can differentiate into eosinophils and basophils ex vivo (Maroz et al. 2014), although it should be noted that basophilic granules have been identified within the TMD blasts.

It has been estimated that ∼10% of infants with DS have TMD based on visible blasts in the peripheral blood and bone marrow (Zipursky et al. 1997). A more recent study has shown that ∼30% of infants with DS have detectable GATA1 mutation (Roberts et al. 2013). The median age at diagnosis is 6.5 d (Maroz et al. 2014). As both T21 and GATA1 mutations occur in utero, it is not a surprise that signs of TMD present at the fetal or neonatal stages.

A fascinating question is why TMD spontaneously resolves in the majority of patients. It is likely that the switch from fetal liver to bone marrow hematopoiesis is involved, with the loss of the permissive environment needed to maintain the TMD clones (Gamis and Hilden 2002). Two elegant studies suggest that differences in IGF (Klusmann et al. 2010a) or interferon signaling (Woo et al. 2013) may account for the remission of TMD.

Data suggest that 10%–20% of newborns with TMD progress to AMKL within four years (Maloney et al. 2015). The median age at diagnosis is 1.7–1.8 yr (Gamis et al. 2003; Sorrell et al. 2012). Outcomes for children with AMKL who have DS are higher than those without DS, with 5-yr overall survival rates of >80% (Sorrell et al. 2012). Taub and colleagues showed that this was associated with higher rates of ara-C metabolism in T21 cells, as indicated by elevated levels of the intracellular ara-C metabolite, ara-CTP (Taub et al. 1996). A subsequent study suggested that the chromosome 21 genes involved in the altered metabolism include cystathionine-β-synthase (CBS) and superoxide dismutase 1 (SOD1) (Taub et al. 1999). When children with DS are diagnosed at 2 yr of age or older, they tend to have higher rates of relapse and worse survival (Gamis et al. 2003). In children older than 4 yr, the disease typically lacks GATA1 mutations and is phenotypically distinct from DS-AMKL. This suggests that T21/GATA1 mutant clones lose their propensity for transformation with time.

The cell of origin in AMKL remains unclear. Evidence suggests that other subtypes of AML originate from a population of preleukemic HSCs, which harbor early founder mutations and undergo clonal evolution to produce HSPCs that are transformed on the accumulation of further mutations (Jan et al. 2012; Corces-Zimmerman et al. 2014; Shlush et al. 2014). These leukemic stem cells (LSCs) are capable of long-term self-renewal and disease initiation and display increased resistance to many therapeutic approaches (Thomas and Majeti 2017). Leukemia blasts originating from transformed HSCs are more aggressive than those arising from myeloid lineage-committed progenitor cells such as common myeloid progenitors (CMPs) and granulocyte-monocyte progenitors (GMPs) (George et al. 2016). It is possible that AMKL stem cells may arise directly from preleukemic HSCs, as several lines of evidence suggest that megakaryocytes may be derived from HSCs rather than from an MEP or other lineage-committed progenitor (Woolthuis and Park 2016). With respect to DS-AMKL, studies by Orkin and colleagues have suggested that there may be a unique hematopoietic progenitor cell in fetal livers with T21; this cell may be highly susceptible to GATA1 mutations, potentially explaining the strong link between the two genetic alterations (Li et al. 2005). Further studies are necessary to precisely define the cell of origin for all the subtypes of AMKL.

ACQUISITION OF TRANSFORMING MUTATIONS

Although GATA1 mutations are seen at both the TMD and AMKL stages, progression from TMD to AMKL is associated with the acquisition of at least one additional mutation (Yoshida et al. 2013). These advanced mutations occur in multiple genes involved in multiple epigenetic or signaling pathways. In particular, more than half of DS-AMKL patients, but few if any TMD patients, harbor mutations in the cohesin complex (Yoshida et al. 2013). Cohesin, a ring-shaped complex consisting of four core subunits, is responsible for tethering sister chromatids together during mitosis, organizing chromatin within topologically active domains, and facilitating promoter–enhancer interactions within these domains to regulate gene expression (Losada 2014). Heterozygous mutations in cohesin subunit genes are also found in other subtypes of AML, albeit at lower frequencies (∼10%), occur early on during leukemic development, and are mutually exclusive across subunits (Fisher et al. 2017b). Multiple in vitro and in vivo experiments have found that cohesin haploinsufficiency alone leads to a disease that resembles a myeloproliferative neoplasm but does not trigger leukemic transformation unless combined with other oncogenes (Mullenders et al. 2015; Viny et al. 2015). Furthermore, knocking down one of the core cohesin subunits or overexpressing a mutant allele imparts alterations in HSPC populations, leads to induction of stem cell genes, and causes global changes in chromatin accessibility (Mazumdar et al. 2015; Mullenders et al. 2015; Viny et al. 2015; Galeev et al. 2016). Altered cohesin complexes exert these changes in part because they fail to recruit the PRC2 complex, which deposits the repressive H3K27me3 histone mark, leading to derepression of stem cell genes Hoxa7/9 (Fisher et al. 2017a).

Cohesin mutations likely cooperate with increased expression of chromosome 21 genes and the Gata1s leukemic isoform in the context of DS-AMKL (Fig. 2). Of note, cohesin disruption has been shown to increase expression of both RUNX1 and ERG in disomic cells, and it is possible that this effect is amplified when three copies of each gene are present cells with T21 (Mazumdar et al. 2015; Fisher et al. 2017a). Cells expressing cohesin mutants also tend to have more open, accessible chromatin surrounding the binding motifs of RUNX1 and ETS family transcription factors such as ERG, and RUNX1 displayed increased global chromatin occupancy (Mazumdar et al. 2015). Additionally, whereas expression of cohesin mutant alleles in human blood cells caused an increase in the population of CD34⁺ HSPCs, knockdown of either RUNX1 or ERG reversed this effect, indicating that these genes play a role in the aberrant stem cell phenotype produced by cohesin loss (Mazumdar et al. 2015). Finally, reducing the number of functional cohesin complexes may also cooperate with Gata1s, as cohesin attenuation leads to increased expression of GATA1 and GATA2 and altered accessibility of GATA motifs (Mazumdar et al. 2015; Mullenders et al. 2015; Viny et al. 2015). However, it remains unclear how alterations in cohesin would affect the activity of Gata1s, which displays altered chromatin binding compared with the full-length protein (Chlon et al. 2015) Additional studies are needed to determine how trisomy 21 genes, Gata1s, and cohesin mutations cooperate at the molecular level to lead to DS-AMKL.

Figure 2.

Model of cooperation among trisomy 21 (T21), GATA1 mutations, and cohesin mutations. Trisomy and subsequent overexpression of chromosome 21 genes leads to enhanced self-renewal of GATA1 mutant cells and increased expression of the Gata1s isoform. The addition of loss-of-functional mutations in cohesin complex components further increases expression of chromosome 21 genes as well as GATA1. It also is associated with increased chromatin occupancy of RUNX1 and an altered chromatin accessibility of RUNX, ETS, and GATA binding motifs.

Mutations in the PRC2 complex member EZH2 were also discovered in a third of DS-AMKL patients, but no patients with TMD (Yoshida et al. 2013). EZH2 has been shown to cooperate with GATA1 to repress HSPC genes, indicating that in a Gata1s context, EZH2 mutations might synergize with Gata1s to produce further derepression of key growth regulatory genes (Gruber and Downing 2015). Other frequently mutated genes in DS-AMKL and TMD include members of JAK-STAT, MAPK/PI3K, and WNT signaling pathways, including JAK2, JAK3, MPL, FLT3, and APC. Notably, MPL mutations in the context of TS21 and Gata1s were sufficient to induce megakaryocytic leukemia in mice (Malinge et al. 2012) Although it is unclear how these mutations may be contributing to the transformation to DS-AMKL, it is notable that these signaling pathways have been linked to up-regulation of MYC in other contexts (Malinge et al. 2008; Nikolaev et al. 2013; Yoshida et al. 2013). Furthermore, disruptions in JAK/STAT and RAS pathways have been shown to contribute to increased numbers of megakaryocyte progenitors and cytokine-independent growth (Gruber and Downing 2015). Mutations in RAS and JAK pathway genes tended to occur late in disease progression and may not exist in all AMKL clones, suggesting that these mutations are less important for disease initiation (Yoshida et al. 2013).

AMKL IN INDIVIDUALS WITHOUT DS

AMKL in children without DS is driven by different genetic pathways than is DS-AMKL. Non-DS-AMKL cells are much more likely to have copy number alterations, contain chromosomal rearrangements, and express disease-driving fusion proteins (Radtke et al. 2009). The observed translocations and the functions of their component genes are detailed in Table 1.

Table 1.

Common fusion events in non-DS-AMKL

Translocation	Disrupted genes	WT protein function	Fusion protein function	References
t(1;22)(p13;q13)	RBM15 (OTT)	Interacts with the RBPJ transcription factor; induces RBPJ-mediated Notch signaling in myeloid progenitor cells but represses it in other cell types; conditional deletion in an in vivo model leads to increased numbers of HSPC, myeloid, and megakaryocyte cells; required for stress hematopoiesis; controls RNA splicing	Interacts with RBPJ to activate Notch signaling; human AMKL cells with the translocation have increased levels of RBPJ target genes; leads to constitutive SRF signaling; disrupts hematopoiesis and induces AMKL in vivo	Mercher et al. 2001; Miralles et al. 2003; Ma et al. 2007; Raffel et al. 2007; Descot et al. 2008; Cheng et al. 2009; Gilles et al. 2009; Mercher et al. 2009; Xiao et al. 2012; Smith et al. 2013; Zhang et al. 2015
	MKL1 (MAL)	Rho-actin signaling leads to its nuclear localization, where it activates serum response factor (SRF) transcriptional activity; up-regulated during megakaryocyte differentiation; its loss disrupts normal megakaryocyte differentiation and migration
inv(16) (p13.3q24.3)	CBFA2T3	Acts as transcriptional corepressor; necessary for HSPC proliferation; its loss results in increased numbers of granulocyte/macrophage cells and decreased numbers of MEPs	Expression in mouse cells leads to increased self-renewal; acts as transcriptional activator and up-regulates hedgehog, BMP, and JAK-STAT signaling; cells expressing the fusion are dependent on growth factors and do not induce leukemia in vivo, indicating that additional mutations may be necessary for transformation	Chyla et al. 2008; Gruber et al. 2012; Thiollier et al. 2012; Gruber and Downing 2015
	GLIS2	Member of Hedgehog signaling pathway, not normally expressed in hematopoietic cells
t(11;15)(p15;q35)	NUP98	Does not have DNA-binding domains but is able to interact with the HOXA7 and -A9 promoter; recruits CREBBP/p300 to induce histone acetylation and gene expression	May accumulate in the nucleus and result in aberrant transcription; blocks PRC2 binding and leads to derepression of PRC2 targets; up-regulates HOXA and HOXB genes, which block myeloid differentiation and maintain HSPC state; induces AML in vivo	Wang et al. 2009; Hollink et al. 2011; Thiollier et al. 2012; de Rooij et al. 2013
	JARID1A (KDM5A)	Demethylates H3K4, interacts with Rb and regulates expression of HOX genes

DS, Down syndrome; AMKL, acute megakaryoblastic leukemia; WT, wild-type; HPSC, hematopoietic stem and progenitor cell; SRF,serum reponse factor; MEP, megakaryocyte-erythroid progenitor; AML, acute myeloid leukemia.

One of the most common chromosomal translocations seen in the disease is t(1;22)(p13;q13), which results in formation of a chimeric oncogene between RNA-binding motif protein-15 (RBM15; also known as OTT) and megakaryoblastic meukemia-1 (MKL-1; also known as MAL) (Mercher et al. 2001). This fusion protein is common in infant AMKL (Mercher et al. 2002) and found in ∼12% of overall pediatric megakaryocytic leukemia cases (de Rooij et al. 2016). RBM15-MKL1 disrupts normal transcription and signaling in hematopoietic cells (Table 1). It likely cooperates with other mutations to induce leukemic transformation. In mice, expression of the fusion alone gave rise to leukemia with low penetrance and long latency (Mercher et al. 2009). However, it induced AMKL at a much higher penetrance when combined with a mutation in MPL (Mercher et al. 2009). Of note, mutations and altered expression levels of matrix metalloproteinase 8 (MMP8) have also been identified in AMKL patients with this translocation (Mercher et al. 2009; Kim et al. 2014). AMKL cases in children without DS who harbor this fusion have a better prognosis that those with other chromosomal alterations (de Rooij et al. 2016).

Another fusion protein often seen in pediatric non-DS-AMKL is CBFA2T3-GLIS2 (also known as ETO2-GLIS2), caused by the inv(16)(p13.3q24.3) rearrangement (Gruber et al. 2012; Thiollier et al. 2012). This translocation is observed in 12%–30% of pediatric AMKL cases (Gruber et al. 2012; Thiollier et al. 2012; de Rooij et al. 2016). The presence of this chromosomal aberration is associated with a worse overall and reduced event-free survival (de Rooij et al. 2016). CBFA2T3-GLIS2 has not been identified in adult AMKL cases, although it does occur in other pediatric AML subtypes (Gruber et al. 2012; Masetti et al. 2013). A recent study found that the fusion protein promotes self-renewal of hematopoietic progenitors by up-regulation and interaction with ERG at enhancer elements (Thirant et al. 2017). Furthermore, disruption of the transcriptional activity of the fusion by blocking its oligomerization promoted megakaryocytic differentiation.

The nuclear pore complex member NUP98 is fused to JARID1A (also known as KDM5A) in 8%–12% of pediatric AMKL cases (Gruber et al. 2012; de Rooij et al. 2016, 2017). This fusion results from the t(11;15)(p15;q35) translocation and likely promotes leukemogenesis through its aberrant up-regulation of HOX genes. It is not a statistically significant prognostic factor in AMKL, although patients with this fusion trend to show lower survival rates than other fusions (de Rooij et al. 2013, 2016, 2017).

Other recurring fusion events include rearrangements in the KMT2A gene in 9%–17% of cases, HOX gene fusions in 15% of patients, and MLL rearrangements in 10% of non-DS-AMKL cases (de Rooij et al. 2013, 2016, 2017). Other rare translocations identified include MN1-FLI1, GRB10-SDK1, C8orf76-HOXA11AS, FUS-ERG, HLXB9-ETV6, RUNX1-CBFA2T3, BCR-ABL1, MAP2K2-AF10, and THRAP-SH3BP2 (Gruber et al. 2012; Thiollier et al. 2012; de Rooij et al. 2013). It is notable that many of these genes play a role in megakaryopoiesis and leukemia. These translocations are nearly always mutually exclusive, although one case was reported to have both CBFA2T3-GLIS2 and reciprocal translocations resulting in both THRAP3-SH3BP2 and SH3BP2-THRAP3 products (Thiollier et al. 2012). The role of these rare fusion events in disease progression is largely unclear. Some are predicted to produce a functional protein and result in a gain or alteration of function of the component genes, whereas others may produce noncoding RNA and induce disease through loss of a tumor suppressive transcript (de Rooij et al. 2017). Notably, the fusion oncogenes seen in non-DS-AMKL have not been detected in AMKL patients with DS (Yoshida et al. 2013).

Several cooperating mutations have been discovered in AMKL, and these are postulated to drive disease progression. For example JAK2, JAK3, and NRAS mutations are also found in DS-AMKL (Gruber et al. 2012; Yoshida et al. 2013). Interestingly, some mutations seem to recur with certain fusion events. For example, nearly all NUP98-KDM5a-expressing cells also have mutated RB1, KMT2A rearrangements frequently co-occur with RAS mutations, and HOX fusion events are associated with mutations in MPL (de Rooij et al. 2017).

Interestingly, many cases of non-DS-AMKL that do not contain oncogenic fusions may be caused by similar genetic pathways as DS-AMKL. For example, AMKL patients without DS have sometimes been found to have acquired T21, increased copy number of Down syndrome critical region (DSCR) genes on chromosome 21, or GATA1 mutations (Yoshida et al. 2013). In one small cohort of non-DS-AMKL patients for which copy number variation data were available, half contained amplification of the DSCR (Gruber et al. 2012). Furthermore, in pediatric non-DS-AMKL cases in which no fusion protein was found, half the patients had GATA1 mutations (de Rooij et al. 2017). Cohesin and CTCF, which are mutated in high frequency in DS-AMKL, are altered in ∼20% of all pediatric non-DS-AMKL patients; these mutations are more likely to occur in non-DS-AMKL patients with mutant GATA1 (de Rooij et al. 2017). The cohesin pathway may also be disrupted through fusion events in these patients. For example, NIPBL, a member of the cohesin loading complex, is occasionally fused to HOXA9 or –B9. Furthermore, the core cohesin complex member STAG2 is found in fusion events that are predicted to result in loss of function of the full-length protein in 3.5% of cases of non-DS-AMKL (Thiollier et al. 2012; de Rooij et al. 2017). Interestingly, the non-DS-AMKL patients carrying GATA1 mutations tend to have better overall and event-free survival, mirroring the improved outcome of DS-AMKL patients (de Rooij et al. 2017).

Finally, in addition to rare cases in children, AMKL is seen in 1%–2% of adult AML; these cases have an extremely poor prognosis (Tallman et al. 2000). Although adult AMKL generally lacks chromosomal translocations, it is characterized by mutations in TP53, DNMT3A, RB1, and genes in the cohesin, splicing factor, and ASXL families (de Rooij et al. 2017). Additional studies on the genetics of adult AMKL are needed to better understand its basis and develop novel targeted therapies.

Given the overall poor outcomes for AMKL, new treatments are desperately needed. This also applies to DS-AMKL, as although the outcome is favorable, patients suffer from severe side effects from the therapy. Recently there has been considerable interest in leveraging differentiation therapy in cancer. With respect to AMKL, inhibition of aurora kinase A has been shown to lead to polyploidization and partial differentiation of AMKL blasts in vitro and in vivo, and to impart a significant survival advantage in animal models of the disease (Thiollier et al. 2012; Wen et al. 2012). Furthermore, a phase I study of the AURKA inhibitor alisertib revealed that the drug led to improved megakaryopoiesis in patients with myelofibrosis, a disease that is characterized by an accumulation of atypical megakaryocytes (Gangat et al. 2019). Further clinical studies are needed to assess the activity of AURKA inhibition in AMKL.

CONCLUDING REMARKS

AMKL is a rare subtype of acute myeloid leukemia in adults, but more common in pediatric patients, especially those with DS. In the recent past, a great deal has been learned about the genetic drivers of pediatric AMKL, which include GATA1 mutations in DS-AMKL and a number of chromosome translocations in non-DS-AMKL, most notably RBM15-MKL1 and CBFA2T3-GLIS2. Nevertheless, there are many unanswered questions: Why are children with DS at such an elevated risk for development of this malignancy? How does trisomy 21 contribute to the initiation of TMD? How do mutations in genes such as CTCF and RAD21 contribute to evolution of TMD to AMKL? Why are GATA1 mutations so common in children with DS, and why are there so many different translocations in non-DS-AMKL? Answers to these questions will shed light on the initiation and progression of AMKL and may lead to the development of new, targeted therapies for this fascinating form of AML.