Acute megakaryoblastic leukemia (AMKL) is a rare but life-threatening hematological malignancy characterized by clonal megakaryoblastic proliferation and impaired differentiation of these cells. AMKL was first described in 1931 and included in the French-American-British (FAB) classification as acute myeloid leukemia (AML) the M7 subtype (De Marchi et al., 2019). In the 2016 World Health Organization (WHO) revision of myeloid neoplasms, AMKL in Down syndrome (DS) (also named as myeloid leukemia associated with DS, ML-DS) was included in myeloid proliferations associated with DS, while AMKL in non-Down syndrome (non-DS) patients was included in the category of acute myeloid leukemia not otherwise specified (AML NOS) (Arber et al., 2016). This categorization remains similar in the latest 2022 WHO and International Consensus classifications (ICC) (Arber et al., 2022; Khoury et al., 2022). However, cases with KMT2A (lysine methyltransferase 2A) and NUP98 (nucleoporin 98) rearrangements now form independent subgroups of AML with defining genetic abnormalities (that also include other non-AMKL phenotypes), and cases with CBFA2T3 (CBFA2/RUNX1 partner transcriptional co-repressor 3)::GLIS2 (GLIS family zinc finger 2) and RBM15 (RNA binding motif protein 15)::MKL1 (megakaryoblastic leukemia 1) fusions belong to a subgroup of AML with other defined genetic alterations. Other cases of AMKL lacking defined genetic alterations remain defined by differentiation, which requires blasts to express at least one of the platelet glycoproteins: CD41 (glycoprotein IIb), CD61 (glycoprotein IIIa), or CD42b (glycoprotein Ib). Myeloid proliferations associated with DS form a subtype of myeloid neoplasms associated with germline predisposition (Arber et al., 2022; Khoury et al., 2022).

AMKL has a bimodal age distribution, so it can be grouped into adult and pediatric. AMKL is very rare in adults accounting for only 1% of all AML (Wen et al., 2011), but in children, AMKL makes up 4%–15% of newly diagnosed AML cases (de Rooij et al., 2017). AMKL is most frequent in young children with DS (median age at diagnosis 1–1.8 years), accounting for the vast majority of AML in DS (∼70%) (Bhatnagar et al., 2016; Masetti et al., 2019b). By contrast, in non-DS patients, AMKL comprises 5%–10% of AML with a slightly older median age at diagnosis ranging from 1.6 to 1.8 years (de Rooij et al., 2016; Maarouf et al., 2019). Genomic studies highlighted the heterogeneity of pediatric AMKL with the specific alterations characterizing different disease subgroups. ML-DS is driven by overexpressed genes and micro-RNAs located on chromosome 21, GATA1 (GATA Binding Protein 1) mutations, and a range of other somatic mutations, in particular those affecting cohesin molecules, signaling pathways, epigenetic regulators, and hematopoietic transcription factors (Boucher et al., 2021; de Castro et al., 2021; Grimm et al., 2021). In contrast, non-DS AMKL is characterized by recurrent chromosomal translocations, complex karyotype, and DNA copy number abnormalities (de Rooij et al., 2017). Children with non-DS AMKL show more heterogeneity and more chromosomal abnormalities than ML-DS. Strikingly, recurrent oncogene fusions are found in more than 70% of non-DS pediatric AMKL, including CBFA2T3::GLIS2, KMT2A rearrangements, NUP98::KDM5A (lysine-specific demethylase 5A), RBM15::MKL1, and HOX (homeobox) gene rearrangements (de Rooij et al., 2016; de Rooij et al., 2017). For completeness, AMKL in adults is more often secondary, preceded by an antecedent hematologic disorder or exposure to chemotherapy or radiotherapy (Dastugue et al., 2002). Adults with AMKL have different cytogenetic and molecular features compared to pediatric patients, showing a diversity of chromosomal abnormalities, the lack of recurrent genetic fusions, and the presence of highly recurrent mutations in genes encoding TP53 (tumor protein p53), cohesins, splicing factors, and epigenetic regulators such as ASXL (additional sex combs-like), DNMT3A (DNA methyltransferase 3 alpha), and TET2 (tet methylcytosine dioxygenase 2) (Dastugue et al., 2002; de Rooij et al., 2017). Cytogenetic and molecular aberrations influence the outcomes of pediatric AMKL subgroups. ML-DS has an excellent prognosis even with low-dose induction chemotherapy, while the outcomes of non-DS AMKL remain unsatisfactory despite improved diagnostics, intensified treatment protocols, and advanced supportive care (de Rooij et al., 2017; Wang et al., 2021). Non-DS children carrying CBFA2T3::GLIS2, KMT2A rearrangements and NUP98::KDM5A have a very poor prognosis (de Rooij et al., 2016; de Rooij et al., 2017). Molecular features of ML-DS have been recently reviewed, including by our group (Boucher et al., 2021; de Castro et al., 2021; Grimm et al., 2021; Li and Kalev-Zylinska, 2022). Here, we provide an updated summary of molecular abnormalities and emerging novel therapeutic strategies in pediatric non-DS AMKL.

Pediatric non-DS AMKL (also referred to in this review as non-DS AMKL for simplicity) is a highly molecularly heterogeneous disease characterized by recurrent and mutually exclusive genetic aberrations, including CBFA2T3::GLIS2 (18%–27%), KMT2A rearrangements (KMT2Ar) (7%–17%), NUP98::KDM5A (11%–12%), RBM15::MKL1 (10%–11%), and HOX rearrangements (HOXr) (14%–15%) (Figure 1A) (de Rooij et al., 2016; de Rooij et al., 2017; Masetti et al., 2019b). A further 4% of patients have non-recurrent fusions engaging hematopoietic transcription factors and epigenetic regulators, e.g., MN1 (meningioma [disrupted in balanced translocation] 1)::FLI1 (friend leukemia integration 1 transcription factor), BCR (breakpoint cluster region protein):ABL1 (ABL1, tyrosine-protein kinase ABL1), and MAP2K2 (mitogen-activated protein kinase 2)::MLLT10 (myeloid/lymphoid or mixed-lineage leukemia translocated to 10) (de Rooij et al., 2017). Molecular alterations are crucial for risk stratification and tailored treatment. Subgroups carrying CBFA2T3::GLIS2, KMT2Ar, and NUP98::KDM5A are associated with poor prognosis and a high risk of disease recurrence (Figure 1B). Thus, intensive chemotherapy and hematopoietic stem cell transplantation (HSCT) are often used in first remission for these patients (de Rooij et al., 2016; de Rooij et al., 2017; Masetti et al., 2019b). In contrast, cases with RBM15::MKL1, HOXr, non-recurrent fusions, and unknown drivers have an intermediate prognosis and are treated with standard chemotherapy (Figure 1B). This review compiled molecular changes reported in pediatric non-DS AMKL (Tables 1, 2). Ongoing molecular characterization of AMKL is needed to guide future mechanistic studies into disease pathogenesis and provide essential clues on developing new targeted therapies.

FIGURE 1. Distribution of distinct molecular subgroups and their prognostic impact in pediatric non-DS AMKL. (A) Distribution of distinct molecular subgroups in pediatric non-DS AMKL. Pediatric non-DS AMKL is characterized by recurrent and mutually exclusive chromosomal abnormalities, primarily including CBFA2T3::GLIS2 (18%–27%), KMT2A rearrangements (KMT2Ar) (7%–17%), NUP98::KDM5A (11%–12%), RBM15::MKL1 (10%–11%), and HOX rearrangements (HOXr) (14%–15%) (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017; Lalonde et al., 2021). Further 4% of patients have non-recurrent fusions engaging hematopoietic transcription factors and epigenetic regulators (de Rooij et al., 2017). Other cases do not have known oncogenic fusions, but GATA1 mutations are detected in about half of this population (overall found in approximately 10% of pediatric non-DS AMKL) (de Rooij et al., 2017). (B) Prognosis of distinct molecular subgroups in pediatric non-DS AMKL. Based on molecular aberrations, three subgroups (CBFA2T3::GLIS2, KMT2Ar, and NUP98::KDM5A) have an unfavorable prognosis, amounting to nearly half of patients; other three subgroups (RBM15::MKL1, HOXr, non-recurrent fusions and unknown drivers) have an intermediate prognosis, accounting for approximately 40% of patients; and patients with GATA1 mutations represent an excellent prognostic group, making up for roughly 10% of children with non-DS AMKL (de Rooij et al., 2017). Among the unfavorable subsets, CBFA2T3::GLIS2 has the worst outcome, with the 5-year overall survival rate being 14% (de Rooij et al., 2017). In contrast, DS-like AMKL has the best outcome, with the 5-year overall survival rate being 100% (de Rooij et al., 2017).

TABLE 1. Key features of distinct molecular subgroups of pediatric non-DS AMKL.

TABLE 2. Gene signatures associated with distinct molecular subgroups of pediatric non-DS AMKL.

The inversion inv(16)(p13.3q24.3) (generating the CBFA2T3::GLIS2 fusion) is the most frequently, but not exclusively detected in 18%–27% of pediatric non-DS AMKL (Figure 1) (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017). A newer study suggested that this lesion may be even more common, as it was identified in 4 of 6 children with non-DS AMKL (Lalonde et al., 2021). CBFA2T3::GLIS2 is not specific to AMKL and occurs in diverse morphologic subtypes of de novo AML (except for acute promyelocytic leukemia (APL) and leukemia with erythroid differentiation), overall found in approximately 8.4% of cytogenetically normal AML cases (Masetti et al., 2013a; Smith et al., 2020). The inv(16)(p13.3q24.3) fuses the transcriptional repressor CBFA2T3 with the zinc finger DNA-binding transcription factor GLIS2. The prevalence of CBFA2T3::GLIS2 appears higher in African American children, accounting for about 29% of patients with this fusion in the AAML0531 and AAML1031 cohorts (Smith et al., 2020). CBFA2T3::GLIS2 characterizes an extremely aggressive leukemic subgroup with a grim prognosis across all differentiation subtypes (overall survival rates 14%–40%) (Gruber and Downing, 2015; de Rooij et al., 2016; de Rooij et al., 2017; Smith et al., 2020). Among CBFA2T3::GLIS2 AMKL, approximately 40% of patients also carry a DHH (desert hedgehog):RHEBL1 (Ras homologue enriched in brain like 1) fusion (Masetti et al., 2013b; Jetten, 2019). The co-occurrence of these two fusions has an even poorer prognosis than the CBFA2T3::GLIS2 alone (Masetti et al., 2013b; Jetten, 2019). Children with CBFA2T3::GLIS2 AML (including AMKL) tend to be younger (median age of onset ∼1.5 years) and display distinct clinical and laboratory features, including stronger expression of CD56 (neural cell adhesion molecule, NCAM) and a more frequent extramedullary involvement compared to other AML (Masetti et al., 2013a; de Rooij et al., 2016; Hara et al., 2017; Smith et al., 2020; Zangrando et al., 2021). Standard karyotyping cannot identify the CBFA2T3::GLIS2, because of its cryptic nature. Immunophenotypic features may suggest the presence of this fusion, as blasts carrying CBFA2T3::GLIS2 have bright CD56 expression with dim-to-negative HLA-DR (human leukocyte antigen DR isotype), CD38, and CD45 referred to as the RAM immunophenotype (Zangrando et al., 2021). The somatic mutational burden of CBFA2T3::GLIS2 subtype is much lower than in other AMKL subgroups (Gruber et al., 2012; Dang et al., 2017; Masetti et al., 2019a). However, mutations in tyrosine kinases such as FLT3 (fms-like tyrosine kinase 3), KIT (tyrosine-protein kinase KIT), RAS (Rat sarcoma), JAK/STAT (Janus kinases/signal transducer and activator of transcription), and transcription factor GATA1 can co-occur (Gruber et al., 2012; de Rooij et al., 2017; Hara et al., 2017). CBFA2T3::GLIS2 can also be accompanied by trisomies of chromosomes 3, 21 and 8, complex karyotype, or hyperdiploidy (Hara et al., 2017; Amano et al., 2020; Smith et al., 2020; Gillam et al., 2022).

CBFA2T3::GLIS2 fuses the 5′ portion of CBFA2T3 (also named as ETO2, MTG16, RUNX1T3 or ZMYND4) in frame with the 3’ region of GLIS2. The majority of chimeric CBFA2T3::GLIS2 fusions (80%) are between exon 11 of CBFA2T3 and exon 3 of GLIS2. Other rare fusion transcripts have also been reported, including between exons 9, 10 or 12 of CBFA2T3 and exons 2 or 3 of GLIS2 (Masetti et al., 2019b; De Marchi et al., 2019; Amano et al., 2020; Smith et al., 2020). The chimeric protein generated by the most common fusion retains most of the functional domains of both proteins with the loss of myeloid, nervy and DEAF-1 domain (MYND) that interacts with nuclear receptor corepressor complex in CBFA2T3, as well as the loss of transactivation domain (TAD) that mediates transcriptional activation of target genes in GLIS2 (Figure 2A). Both CBFA2T3 and GLIS2 participate in leukemic transformation driven by CBFA2T3::GLIS2 (Thirant et al., 2017a).

FIGURE 2. Recurrent gene fusions associated with poor outcomes in pediatric non-DS AMKL. Schematics display the functional domains but are not drawn to scale. The breakpoints are indicated by arrows. (A). Structure of CBFA2T3, GLIS2, and CBFA2T3-GLIS2 proteins. NHR, nervy homology region; MYND, myeloid, nervy, and DEAF-1; TAD, transactivation domain; TRD, transcriptional repression domain; ZF, zinc finger. (B) Structure of KMT2A (MLL) and KMT2Ar (chimeric) proteins. MBD, menin-binding domain; LBD, LEDGF-binding domain; AT, AT hooks; SNL, speckled nuclear localization domains; CXXC, CXXC domain; PHD1/2/3/4, plant homeodomain 1/2/3/4; BD, bromodomain. CS1 and CS2 are the taspase-1 cleavage sites, and FYRN and FYRC are the domains whereby KMT2A-N and KMT2A-C interact after cleavage. TAD, transactivation domain; SET, H3K4 histone methyltransferase domain. KMT2A fusion proteins are caused by chromosomal rearrangements leading to an in-frame fusion between the N-terminus of KMT2A and one of the multiple fusion partners. The breakpoints of KMT2A are located in the region of ∼1400 aa that contains the AT-hook DNA binding motifs and CXXC domains binding unmethylated CpG-containing DNA. Frequencies of the KMT2A fusion partner proteins are shown, with transcription cofactors MLLT3 (AF9) (found in 53%–67% of cases) and MLLT10 (AF10) (found in 21%–47% of cases) being the most common (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017). (C) Structure of NUP98, KDM5A, and NUP98-KDM5A proteins. GLFG, GLFG repeats; GLEBS, Gle2-binding domain; RNA binding, RNA binding domain; APD, autoproteolytic domain; JMJ, Jumonji domain; ARID, AT-rich interactive domain; PHD1/2/3, plant homeodomains 1/2/3.

CBFA2T3 is a member of the RUNX1T1 (runt-related transcription factor 1, translocated) complex. CBFA2T3 belongs to the eight-twenty-one (ETO) family of chromatin-associated proteins, functioning as a master transcriptional coregulator and a Wnt (wingless/integrated) and Notch (neurogenic locus notch homolog protein) signaling suppressor in normal and malignant hematopoiesis (Steinauer et al., 2017; Masetti et al., 2019a; Steinauer et al., 2019; Steinauer et al., 2020; Jakobczyk et al., 2021). CBFA2T3 is expressed in all hematopoietic cells and contains three Nervy homology region (NHR) domains and a MYND domain (Figure 2A). CBFA2T3 participates in hemopoietic stem cell (HSC) self-renewal and differentiation, megakaryocyte-erythrocyte progenitor development, and leukemia stem cell (LSC) expansion (Steinauer et al., 2017; Masetti et al., 2019a; Steinauer et al., 2019; Steinauer et al., 2020; Jakobczyk et al., 2021). Physiological binding partners for CBFA2T3 include transcription factors and chromatin modifiers, and it is generally presumed that CBFA2T3 represses gene transcription through binding to multiple corepressors, such as E-proteins (e.g., E2A basic helix-loop-helix transcription factor), nuclear receptor corepressors (NCOR), and histone deacetylases (HDAC) (Steinauer et al., 2017). CBFA2T3 contributes to transcriptional repression via GATA1-SCL (stem cell leukemia, also known as TAL1) complex during megakaryopoiesis (Schuh et al., 2005; Hamlett et al., 2008). In the megakaryoblastic cell line L8057, CBFA2T3 knockdown enhances megakaryocytic differentiation via elevating gene expression associated with terminal megakaryocytic maturation (Hamlett et al., 2008); whilst CBFA2T3 deletion in a mouse model induces differentiation along the granulocytic-monocytic lineages at the expense of erythroid-megakaryocytic differentiation (Chyla et al., 2008).

GLIS2 is a member of the GLIS subfamily of Krüppel-like zinc finger transcription factors closely related to the GLI and ZIC subfamilies (Scoville et al., 2017). GLIS2 contains a TAD, a transcriptional repression domain (TRD), and a DNA binding domain (five zinc finger motifs) (Figure 2A). GLIS2 function is implicated in the processes of hematopoiesis and leukemogenesis (Jetten, 2019; Pinto and Chetty, 2020). Normally, GLIS2 is not expressed in the differentiating hematopoietic cells (Gruber et al., 2012; Thirant et al., 2017b); however, its expression contributes to HSC repopulation (Holmfeldt et al., 2016). Knockdown of GLIS2 in murine LSK (Lin-SCA1+c-KIT+) cells reduces HSC repopulation, suggesting GLIS2 could regulate HSC engraftment and hematopoietic reconstitution (Holmfeldt et al., 2016). GLIS2 overexpression in MOZ (monocytic leukemia zinc finger protein):TIF2 (transcriptional intermediary factor 2) leukemic cells promotes their differentiation into mature myeloid cells and delays AML development in mice, indicating that GLIS2 inhibits AML initiation by inducing LSC differentiation (Shima et al., 2018).

Although the molecular bases for AMKL transformation by CBFA2T3::GLIS2 are still far from being elucidated, distinct functional properties of this chimeric protein have been illustrated (Masetti et al., 2019a). CBFA2T3::GLIS2 AMKL has a unique gene expression pattern (Gruber et al., 2012; Thirant et al., 2017a; Benbarche et al., 2022). GATA1 downregulation, ERG (ETS-related gene) upregulation, and activation of super enhancer elements may impair megakaryocytic differentiation and increase abnormal self-renewal of leukemic cells in AMKL associated with CBFA2T3::GLIS2 (Thirant et al., 2017a; Benbarche et al., 2022). Super enhancers are clusters of regulatory elements characterized by high intensity of enhancer-related histone tail modifications (Benbarche et al., 2022). Ectopic expression of CBFA2T3::GLIS2 induces the formation of super enhancers, which controls the expression of KIT and platelet-derived growth factor receptor alpha (PDGFRA) involved in leukemic progression (Benbarche et al., 2022). CBFA2T3::GLIS2 positive cases also display higher expression of CD56, a CD56-interacting partner CACNB2 (calcium voltage-gated channel auxiliary subunit beta 2), GABRE (gamma-aminobutyric acid type A receptor subunit epsilon), miR-224 and miR-452 (Smith et al., 2020). Both miR-224 and miR-452 are intronic miRNAs transcribed from the GABRE locus and have numerous mRNA targets involved in immune responses, leukocyte activation, and leukocyte differentiation. Upregulation of other miRNAs (e.g., miR-181b-5p) may be responsible for the reduction in expression of apoptotic and tumor suppressor genes, and simultaneous downregulation of tumor suppressive miRNAs (e.g., miR-196a/b, miR-133a, and miR-199a/b) (Smith et al., 2020). The role of miRNA changes in leukemia initiation by CBFA2T3::GLIS2 is not well defined. Overexpression of miR-181b-5p downregulates apoptotic genes CASP1 (caspase 1) and TRAIL-R2 (TNF-related apoptosis-inducing ligand receptor 2), and the circadian rhythm gene PER2 (period circadian regulator 2), while loss of miR-196b is associated with ITGA2 (integrin subunit alpha 2) and DNMT3B (DNA methyltransferase 3 beta) overexpression (Smith et al., 2020). When highly expressed, ITGA2 and DNMT3B are independent indicators of poor prognosis in AML (Lamba et al., 2018; Lian et al., 2018). CBFA2T3::GLIS2 is also associated with an overexpression of several cell adhesion and cell surface molecules (e.g., extracellular matrix binding, cell adhesion, and integrin binding genes) and signaling pathways (e.g., Hippo, transforming growth factor beta (TGFβ), bone morphogenetic proteins (BMP), JAK/STAT and Hedgehog) (Gruber et al., 2012; Smith et al., 2020). Overall, multiple molecular alterations may contribute to the phenotype of CBFA2T3::GLIS2 leukemia.

Considering the lack of other recurrent mutations in patients carrying CBFA2T3::GLIS2, this fusion probably is the primary genomic alteration. However, the evidence is mixed on whether CBFA2T3::GLIS2 alone is sufficient for leukemic transformation. On the one hand, the introduction of CBFA2T3::GLIS2 into murine bone marrow is insufficient to induce overt leukemia in mice (Dang et al., 2017). On the other hand, the transduction of human CD34+ cord blood stem cells with a lentivirus carrying CBFA2T3::GLIS2 leads to increased proliferation, maturation arrest, and morphologic and immunophenotypic aberrations consistent with AMKL (Le et al., 2022). Further, an inducible transgenic mouse model demonstrated that CBFA2T3::GLIS2 expressed in fetal HSCs leads to a rapid and aggressive AMKL, whereas its expression in adult bone marrow HSCs results in AML (Lopez et al., 2019). In an endothelial cell coculture system, the expression of CBFA2T3::GLIS2 transforms human cord blood hematopoietic stem and progenitor cells (HSPCs) and leads to highly aggressive leukemia in mice (Le et al., 2022).

In summary, cytogenetically cryptic CBFA2T3::GLIS2 fusion is a potent oncogene contributing to malignant transformation and an extremely lethal and treatment-refractory AML. Disease expressing this fusion have unique molecular features with a specific gene expression signature, dysregulated expression of cell adhesion molecules, RTKs, and signaling pathways (e.g., Hippo, TGFβ, BMP, JAK/STAT and Hedgehog). Despite the use of intensive chemotherapy and HSCT, the outcomes of CBFA2T3::GLIS2 leukemia are poor. The mechanism of leukemogenesis associated with CBFA2T3::GLIS2 remains incompletely understood. Therefore, further mechanistic studies are required to inform the development of novel therapies.

Alterations of the KMT2A gene (also known as mixed lineage leukemia 1, MLL1 or MLL gene) located at 11q23.3 are found in both acute lymphoblastic leukemia (ALL) and AML, making up ∼10% of all leukemia cases in all age groups (Winters and Bernt, 2017). KMT2A rearrangements (KMT2Ar) are one of the most common recurrent genetic aberrations found in 15%–25% of all newly diagnosed cases of pediatric AML (Meyer et al., 2006; Meyer et al., 2009; Meyer et al., 2013; Meyer et al., 2018; Hoffmeister et al., 2021; Quessada et al., 2021; Meyer et al., 2023; Yuen et al., 2023). KMT2Ar are particularly prevalent in infant AML, accounting for approximately 30% of children presenting below the age of 2 years. KMT2Ar correlate with monoblastic and monocytic AML subtypes (71.5%–73%) and are much less common in other subtypes (Meyer et al., 2023). In pediatric AMKL, KMT2Ar are reported in 7%–17.4% of patients, with the median age at diagnosis of 1.9 years (range 0.7–12 years) (Chen and Armstrong, 2015; de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017; Masetti et al., 2019b; Maarouf et al., 2019). Numerous KMT2A fusion partners have been identified in children with AMKL, including MLLT1 (myeloid/lymphoid or mixed-lineage leukemia translocated to 1, also known as ENL), AFF1 (AF4/FMR2 family member 1, previously known as AF4), MLLT3 (also known as AF9), MLLT6, MLLT9, MLLT10 (also known as AF10), and SEPT9 (septin 9) (Takita et al., 2009; de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017; Forlenza et al., 2018; Lalonde et al., 2021; Qiu et al., 2022). The two most common KMT2A translocations in AMKL are t(9;11)(p21;q23) (generating KMT2A::MLLT3 fusion) and t(10;11)(q12;q23) (generating KMT2A::MLLT10 fusion) (Figure 2B). Translocation t(9;11)(p21;q23) was found in 8 of 15 (53%), 9 of 14 (64%) and 2 of 3 (66.7%) of patients in three different studies (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017). In the same studies, t(10;11)(q12;q23) was the second most common, detected in 7 of 15 (46.7%), 3 of 14 (21.4%), and 1 of 3 (33.3%) of patients, respectively (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017). The prognosis of KMT2Ar AML varies, heavily relying on its translocation partner. However, in pediatric AMKL KMT2Ar are always regarded as high-risk fusion events linked to a greater risk of relapse and poorer overall survival (de Rooij et al., 2016; de Rooij et al., 2017). KMT2Ar subgroup of pediatric non-DS AMKL tends to have higher white cell count (WCC) and is more common in males, although these differences were not statistically significant in the study by de Rooij et al. (de Rooij et al. 2016). Pediatric KMT2Ar AML was shown to have a higher expression of CD33 compared to KMT2A wild-type AML (Pollard et al., 2021). However, it is unclear whether pediatric KMT2Ar AMKL has the same pattern, so further studies are required. Co-occurring mutations can be found in AMKL associated with KMT2Ar, including in molecules such as NRAS (neuroblastoma RAS viral oncogene homolog), KRAS (Ki-ras2 Kirsten rat sarcoma viral oncogene homolog), PTPN11 (tyrosine-protein phosphatase non-receptor type 11), NF1 (neurofibromatosis type 1), EPOR (erythropoietin receptor), MPL (thrombopoietin receptor), JAK1/2/3, PIK3C2A (phosphatidylinositol-4-phosphate 3-kinase catalytic subunit type 2 alpha), KIT, cohesin and epigenetic regulators, e.g., STAG3 (stromal antigen 3), SETD2 (SET domain containing 2, histone lysine methyltransferase), IDH1 (isocitrate dehydrogenase 1), CREBBP (cyclic adenosine monophosphate response element binding protein binding protein), transcription factors, e.g., GATA1, RNA splicing and regulatory proteins, e.g., U2AF1 (U2 small nuclear RNA auxiliary factor 1) and DDX3X (DEAD-Box helicase 3 X-linked) (de Rooij et al., 2017; Foster et al., 2019). KMT2Ar associated with KRAS mutations confer a particularly poor prognosis (Matsuo et al., 2020). Several additional chromosomal abnormalities can be observed in pediatric AMKL with KMT2Ar, such as trisomy 8, trisomy 21, hyperdiploidy, monosomy 15, and a complex karyotype (de Rooij et al., 2016; de Rooij et al., 2017; Hara et al., 2017). The clinical impact of these co-occurring chromosomal changes remains unclear.

KMT2A represents a key transcription factor and histone-H3 lysine-4 (H3K4) methyltransferase that serves as a master controller for the transcription of critical genes implicated in normal embryonic development and hematopoiesis (Li and Song, 2021; Sugeedha et al., 2021). KMT2A is a large protein (500 kDa, 3696 amino acids) and contains multiple conserved domains, including menin binding motif, LEDGF (lens epithelium-derived growth factor) binding domain, AT-hook (DNA binding) motifs, nuclear localization signals, CXXC domain (nonmethylated CpG DNA binding domain), PHD (plant homeodomain) fingers, bromodomain, FYRN (FY-rich domain N-terminal) domain, taspase 1 cleavage sites, TAD, FYRC (FY-rich domain C-terminal), and SET domains (Li and Song, 2021) (Figure 2B). Interestingly, its N-terminal ∼1400 residues, containing the AT-hook and CXXC domains, act as a transcription factor that recognizes and binds target genes such as HOXA9 and MEIS1 (Meis homeobox 1) (Muntean et al., 2010). The C-terminal SET domain functions as an H3K4 methyltransferase that mediates chromatin modifications associated with epigenetic transcriptional activation (Li and Song, 2021). An oncogenic fusion protein produced by KMT2Ar predominantly consists of the N-terminal DNA-interacting domains of KMT2A (residues 1–∼1400) fused in frame with one of over 100 fusion partners (Li and Song, 2021) (Figure 2B). KMT2A fusions act as oncoproteins in different leukemic cell and animal models, and expression of those fusions can promote proliferation and arrest myeloid differentiation of hematopoietic progenitors, resulting in their accumulation (Slany, 2016; Skucha et al., 2018). It appears that all forms of KMT2A fusion oncoproteins positively regulate the expression of HOX genes and the HOX cofactor MES1 critical for the leukemic transformation of hematopoietic progenitors (Muntean et al., 2010; Chen and Armstrong, 2015). The mechanism of leukemogenesis may therefore rely on the upregulation of target genes by KMT2A fusion proteins and other recruited proteins, e.g., super elongation complexes involving DOT1L (DOT1 like histone lysine methyltransferase) and H3K79 (lysine 79 of histone H3), or polycomb repressive complex 1 (Chen and Armstrong, 2015; Slany, 2016; Li and Song, 2021). KMT2A::MLLT3, the most common form of KMT2Ar in children with non-DS AMKL, is sufficient to induce a myeloproliferative disorder in mice and to generate leukemia in a mouse model using healthy CD34+ cord blood HSPCs (Mulloy et al., 2008; Wei et al., 2008; Stavropoulou et al., 2016). Wild-type MLLT3 is a positive regulator of early erythroid and megakaryocytic cell fates in primitive human cord blood cells, together with GATA1 (Pina et al., 2008). In a mouse model generated using human CD34+ cord blood HSPCs containing KMT2A::MLLT3, the development of a pronounced hypomethylation phenotype is an early event during leukemogenesis (Milan et al., 2022). Leukemia development in this model was associated with the loss of expression of stem-cell associated genes, gain of expression of HOXA9 and MEIS1, and increased expression ratio of S100 (S100 calcium-binding proteins) A8 to A9 proteins. KMT2A fusion proteins, particularly KMT2A::ENL, regulate only a small subset of genes recognized by wild-type KMT2A (e.g., HOXA9, MEIS1, and several other transcription factor genes, e.g., EYA1 (eyes absent homolog 1), SIX1 (SIX homeobox 1) and SIX4 (SIX homeobox 4), highlighting that the transforming capacity of KMT2A fusion may not be limited to HOXA/MEIS1 genes (Wang et al., 2011). Another study showed that IKAROS (IKAROS family zinc finger 1) acts as an essential regulator in KMT2Ar AML by influencing tumor suppressor pathways, immune dysregulation, and changes in cell differentiation (Aubrey et al., 2022). Cyclin-dependent kinase 6 (CDK6), a cell cycle regulator, was also shown to play a role in the development of KMT2Ar AML (Placke et al., 2014). Inhibition of CDK6 by shRNA or pharmacological inhibitor (e.g., PD-0332991) decreases leukemic cell growth and promotes myeloid cell differentiation in cell lines and primary human AML cells harboring different KMT2A translocations. In a mouse model with KMT2A::MLLT3, CDK6 inhibition increases cell differentiation and prolongs mice survival (Placke et al., 2014).

In summary, similar to CBFA2T3::GLIS2, KMT2Ar in pediatric non-DS AMKL drive disease with inferior outcomes, high incidence of treatment non-response and relapse. Children carrying KMT2Ar present with distinct clinical and molecular features, including monocytic differentiation, higher WCC, higher expression of CD33, fewer co-occurring mutations compared with KMT2A wild-type AML, and higher expression of HOXA/MEIS1 genes. However, despite improvements in the molecular characterization of KMT2Ar AML, leukemogenic mechanisms remain incompletely understood, and new therapeutic approaches are needed.

The cytogenetically cryptic translocation t(11;12)(p15;q35) (resulting in NUP98::KDM5A fusion, also known as NUP98::JARID1A or NUP98:: RBP2) was initially identified in pediatric AMKL in 2006 (van Zutven et al., 2006). NUP98::KDM5A fuses NUP98 located on chromosome 11p15 with KDM5A located on the telomeric end of chromosome 12p13.3 (Masetti et al., 2019b). The fusion is found in 2% of pediatric AML overall, primarily in 8%–12% of non-DS AMKL and 12% of infant AML (de Rooij et al., 2013; de Rooij et al., 2016; de Rooij et al., 2017; Masetti et al., 2019b; Hara et al., 2020; Noort et al., 2021). NUP98::KDM5A AML patients have a median age at diagnosis of 3.2 years (ranging from 3 weeks to 18.5 years) (Noort et al., 2021). NUP98::KDM5A can be found in all morphologic subtypes of AML except for APL. Megakaryocytic differentiation was seen in 34% of NUP98::KDM5A AML, monocytic in 21%, and erythroid in 17% of these cases (Noort et al., 2021). The median age of AMKL patients carrying NUP98::KDM5A is lower than in AML overall (1.8–1.9 years, range: 0.7–12 years) (de Rooij et al., 2016; de Rooij et al., 2017; Noort et al., 2021). NUP98::KDM5A confers a dismal prognosis with an overall survival of 33%–36% due to a high incidence of induction failure or relapse (de Rooij et al., 2016; Hara et al., 2020; Noort et al., 2021). Additional chromosomal abnormalities can be found in association with NUP98::KDM5A, in particular trisomy 21, hyperdiploidy, −13/-13q and a complex karyotype (de Rooij et al., 2016).

NUP98 has many different partner genes (over 30) to produce a series of abnormal fusion proteins in several hematopoietic malignancies (including AML, myelodysplastic syndrome, T-ALL, and mixed-phenotype acute leukemia) (Michmerhuizen et al., 2020). KDM5A is the most common fusion partner of NUP98 in AML and AMKL (de Rooij et al., 2017; Cardin et al., 2019), and NSD1 (nuclear receptor) is the second most common (Mercher and Schwaller, 2019). NUP98::KDM5A and NUP98::NSD1 chimeric proteins fuse the N-terminus of NUP98 with the C-terminus of KDM5A (harboring PHD3) or NSD1 (harboring the PHD and SET domains), respectively (Wang et al., 2007; Wang et al., 2009; Hollink et al., 2011; Schmoellerl et al., 2020). NUP98::KDM5A patients have distinct clinical features compared with NUP98::NSD1, presenting at a younger age and showing a lower WCC (Noort et al., 2021). Additional mutations that associate with NUP98::NSD1, particularly those affecting RAS, WT1 and FLT3, are rarely found in association with NUP98::KDM5A (Noort et al., 2021). In contrast, loss of RB1 (Retinoblastoma 1) locus frequently associates with NUP98::KDM5A (90%, n = 9), and GATA1 mutations were present in 2 of 9 NUP98::KDM5A patients with the RB1 loss (de Rooij et al., 2017; Masetti et al., 2019b; Iacobucci et al., 2019; Michmerhuizen et al., 2020). Leukemic cells carrying NUP98::KDM5A show a strong upregulation of HOXA (HOXA5, HOXA9, HOXA10), and HOXB (HOXB2, HOXB3, HOXB4, HOXB5, HOXB6) genes compared with RBM::MKL1 and CBFA2T3::GLIS2 fusions (de Rooij et al., 2013; Noort et al., 2021). Intriguingly, this upregulation of HOX genes is shared with NUP98::NSD1, DEK (DEK proto-oncogene)::NUP214 (nucleoporin 214), and NPM1 (nucleophosmin 1) mutated cases, suggesting HOXA and HOXB overexpression is a common alteration in leukemia development (de Rooij et al., 2013; Noort et al., 2021). Additionally, both NUP98::KDM5A and NUP98::NSD1 AML show upregulation of targets of E2F (E2F transcription factor 1) and FLT3, and downregulation of targets of TP53 and HDAC (Noort et al., 2021). The upregulation of STAT5, NF1, and NOTCH1, and the downregulation of MYC targets were also identified in NUP98::KDM5A cases, but not in NUP98::NSD1 cases (Noort et al., 2021).

NUP98::KDM5A is undetectable with conventional karyotyping, whereas newly developed next-generation sequencing (NGS) technologies can detect it (Mercher and Schwaller, 2019). The most common in-frame fusion is between exon 13 of NUP98 and exon 27 of KDM5A, but an in-frame fusion involving exon 14 of NUP98 has also been described (de Rooij et al., 2013). NUP98:KDM5A protein contains the N-terminal glycine-leucine-phenylalanine-glycine (GLFG) repeats of NUP98 fused to the C-terminal PHD3 finger of KDM5A (Figure 2C) (Gough et al., 2011; Cardin et al., 2019). Both, the GLFC repeat of NUP98 and the PHD3 domain of KDM5A are thought to participate in leukemic transformation (Wang et al., 2009; Gough et al., 2011; Michmerhuizen et al., 2020).

NUP98 is a structural component of a nuclear pore complex responsible for transporting small ions, polypeptides and macromolecules (e.g., RNA and proteins) into and out of the nucleus (Gough et al., 2011; Michmerhuizen et al., 2020). NUP98 also functions as a transcriptional regulator and assists cell cycle progression (Gough et al., 2011). NUP98 protein consists of GLFG repeats, GLEBS (Gle2-binding sequence) binding domain, RNA-binding sites, and autoproteolytic cleavage site (Figure 2C). GLFG repeats are thought to function as docking sites for karyopherins (that support molecular trafficking), CREBBP, p300 (EP300), exportin 1 (XPO1), and the mRNA export factor TAP (Gough et al., 2011; Michmerhuizen et al., 2020). In hematopoietic progenitors, NUP98 regulates H3K4me3 (trimethylation of histone H3 at lysine 4) via binding to promoters adjacent to regions associated with H3K4me3 and via interaction with Wdr82-Set1A/COMPASS (complex of proteins associated with Set1) (Franks et al., 2017). The role of NUP98 in leukemia development depends on its interaction with the PHD3 domain in KDM5A (Wang et al., 2009; Cardin et al., 2019; Zhang et al., 2020).

KDM5A is a histone lysine demethylase that can remove methyl groups from histones H3K4me1/2/3, thus modulating transcriptional activation or repression (Yang et al., 2021). KDM5A is composed of a Jumonji (JMJ) N domain, a Bright/ARID DNA binding domain, a JMJ C domain, and three PHD domains (Figure 2C) (Yang et al., 2021). PHD3 finger is capable of binding to H3K4me1/2/3, with H3K4me3 being the preferred substrate (Yang et al., 2021). KDM5A overexpression in leukemia associates with a poor prognosis (Gale et al., 2016; Yang et al., 2021).

NUP98::KDM5A fusion is sufficient to induce leukemic transformation by altering proliferation, differentiation, maturation, and self-renewal in different cellular and animal models (Wang et al., 2009; Cardin et al., 2019; Domingo-Reines et al., 2022). The bone marrow-derived HSPCs transduced with NUP98::KDM5A fusion show myeloid differentiation arrest and sustained self-renewal (Wang et al., 2009). Overexpression of NUP98::KDM5A in human cord blood stem/progenitor cells results in maturation block and abnormal proliferation with short latency (Cardin et al., 2019). Similarly, in a human embryonic stem cell (hESC) model with doxycycline-regulated NUP98::KDM5A expression, inducible expression of the fusion protein affects progenitor cell production, accompanied by enhanced expression of HOXA gene cluster (Domingo-Reines et al., 2022). Mice transplanted with NUP98::KDM5A bone marrow progenitors develop CD34+CD117+ AML, characterized by transcriptional upregulation of lineage-specific factors (HOXA, GATA3, MEIS1, EYA, PBX1 (pre-B-cell leukemia transcription factor 1), and epigenetic activation of the HOXA gene cluster (Wang et al., 2009). Transfer of human cord blood HSPCs modified to express NUP98::KDM5A in mice results in AML, including AMKL (Cardin et al., 2019). Expression profiling of these synthetic AMKL xenografts closely matches those of AMKL patients (correlation coefficients >0.9). Prominent expression changes include transcriptional upregulation of HOXA, HOXB, MEIS1 and MEIS2, epigenetic activation of HOXB, and overactivation of STAT5A signaling (Cardin et al., 2019). Integrative analysis of transcriptomic and proteomic data from NUP98::KDM5A AMKL models identified cell membrane proteins SELP (selectin P), MPIG6B (megakaryocyte and platelet inhibitory receptor G6b), and NEO1 (neogenin) as novel disease biomarkers (Cardin et al., 2019). Upregulation of JAK-STAT signaling occurs in both synthetic AMKL xenografts and NUP98-rearranged AMKL patient-derived xenografts (Cardin et al., 2019). Like in KMT2Ar AML, CDK6 is highly expressed in NUP98 rearranged AML, representing a critical direct target of NUP98 fusion proteins (Schmoellerl et al., 2020). CDK6 expression is essential for initiating and maintaining AML driven by NUP98 fusion. CDK6 inhibitor palbociclib induces myeloid differentiation, apoptosis and cell cycle arrest in vitro and in vivo (Schmoellerl et al., 2020). Thus, CDK6 inhibition has been proposed as a rational strategy to target NUP98 fusion effects in AML (Schmoellerl et al., 2020).

Although the detailed molecular mechanism of leukemic transformation by NUP98::KDM5A is unclear, it appears that the upregulation of HOX, GATA3, MESI1 and PBX1 transcription factors are essential (Wang et al., 2009; Franks et al., 2017). The N-terminal domain is responsible for recruiting the WDR82-SET1A/B-COMPASS complex to promote H3K4me through which it upregulates gene expression (Franks et al., 2017). The C-terminal PHD3 domain that recognizes H3K4me3 also contributes (Wang et al., 2009). Cell transformation is dependent on the integrity of the PHD3 finger that specifically binds H3K4me2/3. Mutations affecting conserved residues in this domain abrogate H3K4me3 binding, which prevents the binding of NUP98:KDM5A protein to the HOXA9 promoter (Wang et al., 2009).

In summary, NUP98::KDM5A fusion identifies a high-risk group of patients with inferior outcomes. Leukemic cells harboring NUP98::KDM5A have unique molecular features. NUP98:KDM5A is recruited to the promoters of HOX genes associated with H3K4me3/2. This causes transcriptional upregulation of HOXA, GATA3, MEIS1, MEIS2, EYA, and PBX1, and epigenetic activation of the HOXA and HOXB gene clusters. Upregulation of these genes blocks differentiation and maturation, and sustains proliferation and self-renewal of leukemic cells. Nevertheless, the exact molecular mechanism of malignant transformation driven by NUP98::KDM5A is not fully understood. The pathogenetic significance of the RB1 loss in the presence of NUP98::KDM5A is unknown. Novel molecularly guided treatment options are required for this disease.

The t(1;22)(p13;q13) results in a fusion of the RBM15 (the RNA-binding motif protein 15) gene, also named as OTT (one twenty-two) located on chromosome 1p13 to the MKL1 (megakaryoblastic leukemia-1) gene, also named as MAL (megakaryocytic acute leukemia) or MRTFA (myocardin-related transcription factor A) located on chromosome 22q13 (Ma et al., 2001; Raffel et al., 2007). Although the translocation t(1;22)(p13;q13) is rare in pediatric AML (seen in approximately 0.3% of cases) (Quessada et al., 2021), it is almost exclusively seen in infants or young children (age <3 years) with AMKL (10%–15% of pediatric non-DS AMKL) (Carroll et al., 1991; de Rooij et al., 2016; Masetti et al., 2019b; Quessada et al., 2021). The t(1;22)(p13;q13) AMKL is associated with young patient age (median 5–8 months), female prevalence, and intermediate risk (Inaba et al., 2015; de Rooij et al., 2016; Masetti et al., 2019b). Because of the high selectivity of the t(1;22)(p13;q13) for infants with AMKL, the hypothesis was raised that this translocation may arise in utero, causing transformation of a unique developmental stage in HSPCs (Reed et al., 2021). Children with t(1;22)(p13;q13) younger than 6 months do not harbor other cytogenetic abnormalities, while children older than 6 months may have hyperdiploidy with the duplication of der(1)t(1;22) or gains of chromosomes 2, 6, 19, or 21 (de Rooij et al., 2016; Quessada et al., 2021). The RBM15::MKL1 chimeric protein contains all known functional domains of RBM15 and MKL1 (Figure 3A). The molecular mechanisms through which RBM15:MKL1 drives AMKL transformation still need to be better understood. Normal functions of RBM15 and MKL1 provide some clues to this fusion’s roles in the AMKL development.

FIGURE 3. Chimeric fusions associated with favorable outcomes in pediatric non-DS AMKL. (A) Structure of RBM15 (OTT), MKL1 (MAL), and RBM15-MKL1 proteins. RRM, RNA recognition motif; SPOC, Spen paralogue and orthologue C-terminal domain; RPEL, actin-binding motifs with Arg-Pro-X-X-X- Glu-Leu core consensus; B, basic domains; Q, glutamine-rich domain; SAP, homology domain found in SAF-A/B, acinus, PIAS; LZ, leucine-zipper-like domain; TAD, transactivation domain. (B) HOX genes and HOXr identified in pediatric non-DS AMKL (de Rooij et al., 2017). There are 39 HOX genes clustered into four families: HOXA, HOXB, HOXC and HOXD. HOXA7, HOXA9, HOXA10, HOXA11, HOXB8, HOXB9 and HOXD10 are rearranged in AMKL, with HOXA9, HOXA10 and HOXA11 being the most frequently affected (de Rooij et al., 2017).

RBM15 is a crucial regulator of N6-methyladenosine methylation of RNA (Zhao et al., 2022). It contains three RNA recognition motifs (RRMs) and a SPEN paralogue and orthologue C-terminal (SPOC) domain (Figure 3A) (Raffel et al., 2007; Niu et al., 2009; Patil et al., 2016; Jin et al., 2018; Schuschel et al., 2020). RBM15 can regulate RNA splicing and histone modification of MPL critical for the HSC and megakaryocyte function (Xiao et al., 2015). RMB15 modulates several transcription factors associated with megakaryocytic differentiation, including RUNX1 (runt-related transcription factor 1), GATA1, and c-MYC (Niu et al., 2009; Zhang et al., 2015). The depletion of RBM15 in the human megakaryoblastic leukemia cell line Meg-01 enhances the formation of alternatively spliced isoforms of RUNX1a and GATA1s (Zhang et al., 2015). RUNX1a (also known as AML1a), a C-terminally truncated RUNX1 isoform, increases DNA binding and affects target gene transcription, with its overexpression increasing functional HSCs and decreasing hematopoietic differentiation (Davis et al., 2021). GATA1s, an N-terminally truncated GATA1 isoform, plays a major role in transient abnormal myelopoiesis and ML-DS development by promoting megakaryocytic progenitor expansion and disrupting megakaryocytic and erythroid differentiation (Wechsler et al., 2002; Shimizu et al., 2009; Chlon et al., 2015; Banno et al., 2016; Juban et al., 2021). The c-MYC proto-oncogene was discovered as a target of RBM15 during HSC and megakaryocyte development (Niu et al., 2009). RBM15 is also involved in regulating the activity of RBPJ (recombination signal binding protein for immunoglobulin kappa J region), which is an important player in the Notch pathway (Ma et al., 2007). RBM15 overexpression suppresses myeloid differentiation, while its knockdown enhances myeloid differentiation in the myeloid precursor cell line 32DWT18 (Ma et al., 2007). In human umbilical cord blood CD34+ cells, RBM15 knockdown inhibits the maturation of megakaryocytes (Jin et al., 2018). RBM15 deletion is lethal in embryonic mice, while conditional-knockout causes pleiotropic effects in stem cells and progenitors, including megakaryocytic expansion in the bone marrow and spleen (Raffel et al., 2007).

MKL1, a transcriptional coactivator of serum response factor (SRF), has two isoforms consisting of two or three N-terminal RPEL-repeats, basic regions, a glutamine-rich domain, a SAP domain (homology domain found in SAF-A/B, acinus, PIAS), a leucine zipper-like domain, and a TAD domain (Figure 3A) (Scharenberg et al., 2010; Kalita et al., 2012). MKL1 regulates cell morphology, adhesion, migration and differentiation in various cell types, including myeloid cells (Reed et al., 2021; Sprenkeler et al., 2021; Tabuchi and Ihara, 2021). Beyond being a coactivator for SRF, MKL1 modulates certain transcription factors (e.g., SMADs) and impacts histone modifications (Reed et al., 2021). MKL1 knockout mice show partial embryonic lethality and aberrant megakaryopoiesis characterized by increased progenitor numbers and reduced numbers of mature megakaryocytes (Sun et al., 2006; Cheng et al., 2009). The ability of MKL1 to promote megakaryocytic maturation largely depends on the SRF regulatory axis. Knockout of either MKL1 or SRF reduces megakaryocytic maturation in primary cultures, while MKL1 overexpression promotes megakaryopoiesis by augmenting both the genomic associations and activity of SRF (Rahman et al., 2018).

The RBM15::MKL1 fusion protein encompasses all putative functional domains encoded by both genes. Thus the fusion possesses the function of both proteins, including the ability to constitutively activate RBPJ, MKL1-and SRF- dependent target genes (Figure 3A) (Descot et al., 2008; Mercher et al., 2009). In a conditional knockin mouse model, RBM15::MKL1 causes abnormal megakaryopoiesis during embryonic and adult development but rarely generates AMKL (Mercher et al., 2009). However, in combination with a MPL mutation, RBM15::MKL1 leads to rapid cell transformation and a fatal disease with features similar to human AMKL (Mercher et al., 2009). This suggests that cooperating mutations are required to develop leukemia in the presence of RBM15::MKL1. One mechanism through which RBM15::MKL1 induces leukemia in vitro and in vivo is through aberrant binding and activation of RBPJ, which is essential for the differentiation bias toward megakaryocytes and proliferation of leukemic cells (Mercher et al., 2009). RBM15::MKL1 may also exert its oncogenic functions through dysregulation of MKL1 and SRF target genes caused by overactivation of MKL1-and SRF-dependent gene transcription (Descot et al., 2008). RBM15::MKL1 is associated with histone modifications, which may lead to the epigenetic deregulation of genes that control megakaryopoiesis (Lee and Skalnik, 2012). Levels of RBM15::MKL1 expression and its endogenous components may also contribute to leukemogenesis. For example, RBM15::MKL1 overexpression decreases endogenous RBM15 levels and increases endogenous MKL1 expression in the megakaryoblastic leukemia cell line 6133, while RBM15 overexpression reduces the fusion protein expression (Lee and Skalnik, 2012). It seems the N-terminal domain of RBM15 controls endogenous RBM15 expression, but the exact mechanism of this regulation is unknown (Lee and Skalnik, 2012). Relevant hPSC-based human models have been developed to study the impact of RBM15::MKL1. These models recreate AMKL features seen in patients, including overexpression of adhesion molecules CDH2 (cadherin 2) and ITGB1, and several components of Notch signaling (Ayllon et al., 2017).

In summary, RBM15::MKL1 is almost exclusively seen in infants or young children (age <3 years) with AMKL. RBM15::MKL1 possesses the function of both RBM15 and MKL1 proteins, including constitutive activation of RBPJ, MKL1-and SRF- dependent target genes. By itself, RBM15::MKL1 appears insufficient to drive AMKL. However, with a MPL mutation, RBM15::MKL1 causes rapid transformation to AMKL. It remains to be determined how RBM15::MKL1 alters megakaryopoiesis and the mechanism through which cooperating mutations drive leukemogenesis.

HOX rearrangements (HOXr) are seen in 14.9% of pediatric non-DS AMKL and associate with better outcomes (5-year overall survival of 77%) than the aforementioned molecular subgroups (CBFA2T3::GLIS2, KMT2Ar, NUP98::KDM5A, and RBM15::MKL1) (de Rooij et al., 2017). The median age at diagnosis for these patients is 1.5 years (range of approximately 6 months to 2 years) (de Rooij et al., 2017; Masetti et al., 2019b). There are a variety of HOXr in AMKL, including GATA2::HOXA9, GATA2::HOXA10, NIPBL (Drosophila melanogaster Nipped-B)::HOXA9, NIPBL::HOXB9, GATA2::HOXA10, EWSR1 (Ewing Sarcoma breakpoint region 1)::HOXB8, PLEK (pleckstrin)::HOXA11AS (HOXA11 antisense RNA), BMP2K (BMP-2-inducible protein kinase)::HOXD10, EP300::HOXA7, C8orf76 (C8orf76, chromosome 8 open reading frame 76)::HOXA11AS, HOXA11::BZW2 (basic leucine zipper and W2 domains 2), HOXA9::ANGPT1 (angiopoietin 1) and HOXA10AS::CD164 (sialomucin core protein 24) (Figure 3B) (de Rooij et al., 2017; Masetti et al., 2019b). HOXr cause the upregulation of the HOX gene in the fusion and of the adjacent HOX genes (de Rooij et al., 2017). As highlighted in previous sections, overexpression of HOX genes is also associated with KMT2Ar and NUP98::KDM5A, so together, increased expression of HOX genes occurs in approximately half of pediatric non-DS AMKL (de Rooij et al., 2017). Multiple other mutations can be seen in association with HOXr in AMKL, including in NRAS, KRAS, MPL, JAK2, PI3K3R1, PTEN (phosphatase and tensin homolog deleted on chromosome 10), STAG2, CTCF (CCCTC-binding factor), and other genes (e.g., RB1, RUNX1, SETX (senataxin), ATM (ataxia-telangiesctasia mutated), SMARCA2 (SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily A, member 2), NSD1, and TP53) (de Rooij et al., 2017). HOXr AMKL is enriched in MPL mutations (41.7%, n = 12) (de Rooij et al., 2017). Murine bone marrow cells transfected with MPL p.W515L and GATA2::HOXA9 or NIPBL::HOXB9 show growth advantage and increased phosphorylation of JAK2 and STAT5 (de Rooij et al., 2017). Additional cytogenetic aberrations observed in HOXr AMKL include gains or losses of certain chromosomes,