Precision medicine for high-risk gene fusions in pediatric AML: a focus on KMT2A, NUP98, and GLIS2 rearrangements Free

Acute myeloid leukemia (AML) is one of the most difficult to treat cancers in children and adolescents/young adults (AYAs) due to high rates of chemoresistance and relapse. Comprehensive next-generation RNA- and DNA-based sequencing and epigenetic analyses of diagnostic specimens from patients with newly diagnosed AML demonstrate significant differences in the mutational spectrum and biology across the age spectrum.^1-5 Fms-related receptor tyrosine kinase 3 internal tandem duplication (FLT3-ITD) and other somatic mutations (ie, missense, frameshift, nonsense, point mutation, and ITD/partial tandem duplications) in epigenetic-associated genes IDH1, IDH2, DNMT3A, and TET2, occur commonly in adults with AML, many of whom have antecedent myelodysplastic syndrome or clonal hematopoiesis.^6,7 In contrast, pediatric AML is driven primarily by recurrent structural variants¹ (chromosomal translocations and gene fusions) with frequencies of 68% in children <2 years and 54% of children 2 to 10 years of age, which contrasts with frequencies <10% in older adults (>75 years) with AML.⁵ The inverse is true for somatic mutations with an average rate of 2 variants per patient in children <2 years, which are often subclonal and of unclear prognostic significance, and 4.4 mutations per patient in adults >75 years. In young children, oncogenic fusions are oftentimes the only abnormality identified, suggesting they are sufficient to promote and maintain leukemogenesis. Of note, myelodysplastic syndrome in children is uncommon and usually arises in the context of an underlying bone marrow failure with monosomy 7 or other genetic predisposition syndromes.⁸ 

Robust characterization of previously unknown genetic drivers within childhood AML and correlation with end-induction measurable residual disease (MRD) response and clinical outcomes data have refined risk stratification during the past decade.⁹ Allocation of pediatric patients with high-risk AML-associated genetics and/or positive MRD to allogeneic hematopoietic stem cell transplant (HSCT) in first complete remission (CR) has been successful for some cohorts,^10-12 whereas improvement in overall survival (OS) over event-free survival (EFS) has not been clearly demonstrated for children with specific rare genetic subtypes, including CBFA2T3::GLIS2 fusions or NUP98 rearrangements.^1,13,14 

Knowledge of high-risk genetic drivers of childhood AML has not necessarily led to better therapeutic outcomes with the exception of improved survival of patients with FLT3-ITD AML with FLT3 kinase inhibitor addition to chemotherapy.¹⁵ Most structural alterations involve transcription factors, which are generally not directly druggable due to inherent structural disorder, lack of defined small-molecule binding pockets,¹⁶ and/or lack of a therapeutic window in selective targeting the oncogenic fusion without inhibition of essential wild-type protein function.¹⁷ Furthermore, the impact of such fusion oncogenes upon molecular mechanisms of leukemogenesis remains incompletely understood. However, RNA sequencing approaches and preclinical modeling studies have recently identified unique biologic vulnerabilities driven by some childhood AML-associated fusion oncogenes, which has engendered development of targeted therapeutic approaches for “boutique” subgroups with historically dismal outcomes. In this review, we describe the current genetic landscape of 3 of the most lethal childhood AML subtypes harboring KMT2A rearrangements, NUP98 rearrangements, or CBFA2T3::GLIS2 fusions and discuss how detailed molecular and immunophenotypic characterization and functional studies have identified critical leukemogenic dependencies potentially amenable to precision medicine approaches with small-molecule inhibitors and antibody-based or cellular immunotherapies.

KMT2A (histone lysine N-methyltransferase 2A, formerly MLL [mixed lineage leukemia]) is a histone methyltransferase ubiquitously expressed in HSCs that functions as a large multidomain protein in complex with additional proteins.¹⁸ Translocation of the N terminus of KMT2A with the C terminus of 1 of >130 identified genes are frequent oncogenic drivers of pediatric AML, acute lymphoblastic leukemia, and mixed phenotype acute leukemia (Figure 1).¹⁹ The range of KMT2A 3′ fusion partners is more diverse in AML compared with acute lymphoblastic leukemia with MLLT3, MLLT10, MLLT1, MLLT4, and ELL as the most commonly involved genes (Table 1).^19,20,KMT2A rearrangements are particularly prevalent in children with AML <3 years of age, comprising ∼40% of cases,¹ and account for 10% to 15% of AML occurring in older children and AYAs.^1,24 As with other driver fusions, the presence of additional leukemia-associated mutations is uncommon with the exception of subclonal variants in NRAS or KRAS that remain of incompletely elucidated clinical significance.^1,25,26 

KMT2A rearrangements are associated with an overall 5-year EFS of 44% in children with newly diagnosed AML,²⁰ although outcomes vary widely by 3’ fusion partner.^27,28 A retrospective analysis of patients treated on the Children’s Oncology Group (COG) AAML0531 phase 3 clinical trial demonstrated improvements in 5-year EFS with addition of the CD33 antibody-drug conjugate gemtuzumab ozogamicin (GO) to conventional chemotherapy (EFS, 48% vs 29%; GO vs no GO; P = .003). Although differences in OS did not reach significance (63% vs 53%; P = .054), children with KMT2A-rearranged AML who achieved a CR had a lower relapse rate (40% vs 66%, GO vs no GO; P = .001) and improved 5-year disease free survival (57% vs 33%; GO vs no GO; P = .002). Benefit of GO addition to chemotherapy in pediatric patients with newly diagnosed KMT2A-rearranged AML was recently independently validated in the MyeChild 01 phase 3 clinical trial with superiority of 3 GO doses vs 1 dose in induction course 1 (2-year EFS 82% vs 63% and OS 92% vs 80%, respectively).²⁹ Positive results from AAML0531 and other studies led to European Medicines Agency and US Food and Drug Administration approvals of GO for children with relapsed and/or newly diagnosed AML.^27,30-32 

In addition to therapeutic benefit of targeting cell-surface proteins such as CD33, significant attention has also been paid to elucidation of critical intracellular biologic dependencies in KMT2A-rearranged AML potentially amenable to precision medicine therapeutics. Homeobox (HOX) genes are a family of transcription factors, characterized by a homeodomain DNA–binding motif. KMT2A positively regulates HOX family gene expression and along with associated cofactors, including MEIS, is required for normal hematopoietic development.^33,34 HOXA7, HOXA9, and MEIS are constitutively expressed in KMT2A-rearranged leukemia.³⁵ Determining cofactors that contribute to the aberrant gene expression profile of KMT2A-rearranged leukemias has led to the identification of several downstream targets (Figure 2).

Disrupter of telomeric silencing 1-like (DOT1L) is a histone methyltransferase that modifies histone H3 on lysine 79 (H3K79).^36,DOT1L binds to KMT2A fusion proteins as part of a complex.^37,KMT2A-rearranged leukemias require H3K79 methylation to drive oncogenic programs, as loss of DOT1L promotes differentiation and decreases expression of KMT2A fusion gene targets, including HOXA and MEIS.³⁷ Preclinical studies of the DOT1L inhibitor pinometostat demonstrated on-target inhibition of the H3K79 methyltransferase activity of DOT1L via occupation of its S-adenosyl methionine–binding pocket³⁸ and inhibition of KMT2A-rearranged leukemia proliferation.³⁶ However, limited phase 1 clinical trial activity of pinometostat was observed in adults and children with KMT2A-rearranged leukemias despite detection of reduced methylation of KMT2A fusion target genes in patients’ leukemia cells, likely in part due to inadequate drug levels achieved.^39,40 

Another essential oncogenic cofactor for KMT2A-fused leukemia is the tumor suppressor protein menin (encoded by MEN1).⁴¹ The N terminus of KMT2A contains a high-affinity conserved binding motif for menin; this interaction, along with the cofactor LEDGF, mediates binding of the fusion complex to chromatin at HOXA7, HOXA9, and MEIS gene promoter loci for initiation and maintenance of leukemogenesis.³³ Nearly 2 decades ago, preclinical MEN1 genetic deletion studies showed reversal of aberrant HOXA gene expression with ensuing inhibition of leukemia proliferation and restoration of differentiation in KMT2A-rearranged leukemia models.⁴¹ Highly selective pharmacologic inhibitors of menin that disrupt the KMT2A-menin interaction and assembly of KMT2A fusion complex on chromatin have been subsequently developed with robust biologic activity reported.⁴² These drugs downregulate the expression of HOX family genes (eg, HOXA9, HOXA10, and MEIS1), inhibit cell proliferation, and induce differentiation to evoke potent antileukemia activity.^42-51 Given the essential role of KMT2A and menin in normal hematopoiesis, it was initially unclear if menin inhibitors would ultimately be too toxic for successful clinical use, but definitive studies have subsequently demonstrated minimal on-target/off-tumor effects in preclinical models.⁴³ 

Informed by the knowledge that wild-type KMT2A also requires an interaction with menin to maintain normal hematopoeisis, researchers have started to investigate the potential activity of menin inhibition in other AML subtypes with increased HOX gene expression.^33,52,53 The most well-studied example to date is NPM1 (nucleophosmin 1)–mutant AML, which comprises 6% and 13% of AML in children and AYAs and is associated with more favorable clinical responses to chemotherapy and low-risk stratification.^24,54-56 Disruption of the oncogenic menin-KMT2A wild-type interaction with menin inhibitors in NPM1-mutant cells reverses constitutive HOXA and HOXB gene expression, displaces menin from HOXA, HOXB, and MEIS loci, and induces myeloid differentiation.^57,58 

Based upon the remarkable collective preclinical data to date, menin inhibitors are now under early-phase clinical evaluation in adults and children with relapsed/refractory acute leukemias harboring KMT2A rearrangements, NPM1 mutations, NUP98 rearrangements (described in detail below), or other biologically relevant genetic alterations (Table 2).^59,60 Promising single-agent efficacy of revumenib (formerly, SNDX-5613), an oral selective inhibitor of the menin-KMT2A interaction, was recently reported via the AUGMENT-101 phase 1 trial (NCT04065399). Thirty-three percent (15/46) of adults with KMT2A-rearranged and 36% (5/14) with NPM1-mutant AML achieved CR or CR with partial hematologic recovery (CRh) with revumenib monotherapy.^61,62 Pediatric-specific data from the AUGMENT-101 trial showed composite CR (CRc) in 39% (5/13) of treated children, with 3 patients achieving MRD negativity (CRc = CR + CR with incomplete count recovery + CR with incomplete platelet recovery).⁶³ The median time to response was 2.3 months (range, 1.0-3.9). Differentiation syndrome (DS) was reported in 8 children, which resolved with hydroxyurea and/or corticosteroid therapy. Corrected QT interval prolongation (grade ≥2) occurred in 1 child.⁶³ Revumenib was approved by the US Food and Drug Administration in November 2024 for children (≥1 year) and adults with relapsed/refractory KMT2A-rearranged acute leukemias based upon the remarkable clinical outcomes observed in this study. However, a critical observation in the AUGMENT-101 trial was emergence of MEN1 mutations in 39% of patients treated with ≥2 cycles of revumenib monotherapy, which facilitated drug resistance and progressive disease.⁶⁴ Alternative epigenetic mechanisms of menin inhibitor resistance, including loss of the polycomb repressive complex 1.1, have also been reported in preclinical studies and may be overcome with venetoclax cotherapy.⁶⁵ It is not yet known whether combination of revumenib with chemotherapy will decrease resistance mutation frequency or if subsequent treatment with alternative first-generation menin inhibitors will be effective given potential for therapeutic cross-resistance.^66,67 Second-generation menin inhibitors may be needed to overcome this obstacle.

The AUGMENT-102 phase 1/2 clinical trial (NCT05326516) subsequently demonstrated safety and enhanced activity of revumenib with fludarabine and cytarabine (FLA) reinduction chemotherapy in children and some adults with relapsed/refractory KMT2A-rearraged acute leukemias, NPM1-mutant AML, or NUP98-rearranged AML.⁶⁸ The study enrolled 27 patients, including 20 (74%) children and adolescents <18 years. Nine of 18 patients (50%) treated with revumenib at dose level 2 (DL2) achieved CRc, nearly all of which were MRD-negative. The median time to response in the DL2 cohort was one month. Interestingly, fewer severe cytopenias were reported in DL2 than DL1 patients, likely attributable to more rapid remission induction with higher revumenib dosing. QTc prolongation (grade ≥2) was reported in 4 patients (15%). Importantly, no patient in AUGMENT-102 experienced DS, which may be due to effective leukemia debulking with FLA chemotherapy.⁶⁸ Potential occurrence of MEN1 mutations has not yet been reported for the AUGMENT-102 trial. Additional studies are now exploring the efficacy of revumenib in combination with hypomethylating agents (eg, azacitidine, decitabine) and B-cell lymphoma 2 (BCL-2) inhibition (eg, venetoclax). One institutional phase 1 trial reported a 58% CR/CRh rate in 15 of 26 patients with relapsed/refractory AML treated with revumenib, ASTX727 (decitabine/cedazuridine), and venetoclax.⁶⁹ A multi-institutional Leukemia & Lymphoma Society-supported Beat AML consortium phase 1b subtrial also recently reported achievement of 100% CRc in 13 adults with relapsed/refractory AML treated with revumenib, azacitidine, and venetoclax (NCT03013998).⁷⁰ Other studies of revumenib in combination with radiolabeling or multiagent chemotherapy are in progress (NCT05406817, NCT05886049, NCT06222580, and NCT06284486; Table 2).

Ziftomenib (formerly KO-539) is another oral selective menin inhibitor that has demonstrated similarly promising safety and preliminary efficacy data. The KOMET-001 phase 1/2 trial (NCT04067336) enrolled a total of 83 adults with relapsed/refractory KMT2A-rearranged or NPM1-mutant AML and reported CR/CRh in 9 of 36 (25%) patients treated with ziftomenib monotherapy at the recommended phase 2 dose.⁷¹ DS was the most significant dose-limiting toxicity, occurring in 18 of 83 patients (22%). Ziftomenib is now under investigation in combination with intensive (cytarabine and daunomycin) or nonintensive chemotherapy (azacitidine and venetoclax) regimens in adult patients with newly diagnosed or relapsed/refractory KMT2A-rerarranged or NPM1-mutant AML via the KOMET-007 phase 2 trial (NCT05735184) with demonstrated safety and robust interim clinical activity reported.⁷² Pediatric-specific investigation of ziftomenib in combination with FLA chemotherapy in children with relapsed/refractory acute leukemias is also being pursued via the Leukemia & Lymphoma Society Pediatric Acute Leukemia/European Pediatric Leukemia consortium–sponsored APAL2020K phase 1/2 trial (NCT06376162).⁷³ 

Additional selective oral menin inhibitors bleximenib (formerly JNJ-7527661) (NCT04811560),⁷⁴ BMF-219 (NCT05153330), and DSP-5336 (NCT04988555) are also under early-phase clinical investigation in adults with relevant genetic subtypes of relapsed/refractory acute leukemias. Preliminary results of a phase 1b clinical trial in 45 adults treated with bleximenib, azacitidine, and venetoclax reported safety and tolerability of this regimen with CRc achieved in 10 of 21 patients (48%) treated in the efficacy phase (NCT05453903).⁷⁵ An acceptable safety profile and clinical activity of bleximenib combined with more intensive cytarabine and daunorubicin or idarubicin (“7 + 3” regimen) in a separate arm of this trial has also been reported.⁷⁶ 

Rearrangement of the nuclear pore complex gene NUP98 (nucleoporin 98) with 1 of more than thirty 3’ gene partners identified to date occurs in ∼4% of children with AML, which is likely an underrepresentation given the cryptic nature of these fusions frequently not detected by conventional karyotyping.^1,22,77 The most common fusion partners are NSD1 and KDM5A, which are associated with specific pediatric age predilections.^1,78,NUP98::NSD1 and NUP98::KDM5A AML subtypes occur at a median of 10.2 and 2.7 years of age, respectively.²² Regardless of the fusion partner, NUP98-rearranged AML is universally associated with chemotherapy resistance and poor clinical outcomes with 4-year EFS ranging from 10% to 33% despite maximally intensive multiagent chemotherapy and often HSCT in the first CR.^78,79 

The N terminus region of NUP98 contains the intrinsically disordered phenylalanine-glycine/glycine-leucine-phenylalanine-glycine (FG/GLFG) repeats, which function as nuclear transport receptor docking sites and as binding sites for transcriptional co-activators (Figure 1).⁸⁰ In oncogenic NUP98 rearrangements, the NUP98 N terminus bind to the C terminus of a fusion partner, which is frequently enriched in genes involved in epigenetic and transcriptional regulation.^81,82 Because the nuclear pore targeting region lies in the C terminus of NUP98, the fusion construct is not retained in the nuclear pore and instead localizes to the nucleus.⁸³ Despite various fusion partners, many NUP98 rearrangements target common genes, which suggests the NUP98 N terminus is involved in DNA binding. Identification of protein complexes with DNA-binding capacities that interact with NUP98 fusions has uncovered novel mechanisms of leukemogenesis.⁸⁴ 

Transcriptional cofactors and epigenetic modifiers, including CREBBP (CREB binding protein), EP300 (E1A binding protein P300), XPO1 (exportin 1), and KMT2A, bind to the FG/GLFG repeats in the NUP98 N-terminus and drive downstream transcriptional signatures.⁸⁵ Recruitment of lysine acetyltransferases CREBBP and EP300 to the FG/GLFG repeats in NUP98::HOXA9 and NUP98::HOXA10 models is sufficient for leukemia initiation.^85,86 The nuclear transport receptor XPO1 through its interaction with the FG/GLFG NUP98 repeats has been shown to facilitate the recruitment of NUP98::HOXA9 to chromatin, enabling transcription of HOX family genes in mouse embryonic stem cells.⁸⁷ Additional work showed that direct interaction of XPO1 with NUP98::HOXA9 or NUP98::DDX10 fusions in myeloid cells caused nuclear accumulation and increased transcription of NFAT and NFkB, transcription factors exported by XPO1.⁸⁸ Disruption of these interactions may have therapeutical potential for NUP98-rearranged AML. For instance, inhibition of XPO1 with leptomycin B decreased HOXA-C gene expression and reduced nuclear puntae formation.⁸⁷ Treatment with the histone deacetylase inhibitor trichostatin A decreased binding of both XPO1 and NUP98::HOXA9 to chromatin at HOX gene cluster regions.⁸⁷ These studies yield mechanistic insight into the potential use of XPO1 inhibitors, such as selinexor, for the treatment of NUP98-rearranged AML.

The KMT2A complex also binds to GLFG repeats in NUP98::HOXA9 models.⁸⁴ Akin to KMT2A-rearranged AML, dysregulation of HOXA genes and MEIS is common in NUP98-rearranged AML, regardless of the 3’ gene partner.^81,89 Similar to NPM1-mutant AML, increased expression of HOXB genes is observed in NUP98-rearranged AML.^78,84 Binding of NUP98 fusion proteins to KMT2A in a histone-modifying complex results in activation of HOX family and MEIS genes, which appears critical for NUP98-altered leukemia initiation/maintenance based upon results of genetic deletion or overexpression studies.⁸⁴ Studies have attempted to exploit this critical KMT2A dependence of NUP98-rearranged AML via investigation of pharmacologic menin inhibition. Experiments with VTP50469 (a revumenib precursor tool compound) in NUP98-rearranged mouse models showed global loss of menin and displacement of KMT2A and NUP98 fusion proteins from chromatin at critical genes, including MEIS1.⁹⁰ Menin inhibition induced transcriptional and phenotypic changes that promoted differentiation, akin to the observations in KMT2A-rearranged/NPM1-mutant leukemias. Overexpression of MEIS was conversely protective against menin inhibitor–induced cell differentiation in NUP98-rerarranged AML models and preserved leukemogenic phenotypes. Additional studies of NUP98-rearranged, patient-derived xenograft models demonstrated in vivo sensitivity to menin inhibition, although co-occurring mutations in other genes may influence depth of response and require further study.⁹⁰ To date, a small number of patients with relapsed/refractory NUP98-rearranged AML have been treated clinically with menin inhibitors with reports of at least 1 MRD-negative CR achieved with revumenib-based therapy.⁹¹ Additional investigation of ziftomenib-based chemotherapy specifically in this patient population is planned via the pediatric APAL2020K phase 1/2 clinical trial and other studies (Table 2).

Additional work integrating RNA- and chromatin immunoprecipitation–sequencing data from preclinical models of NUP98::NSD1, NUP98::KDM5A, and NUP98::DDX10 AML has also identified CDK6 (cyclin-dependent kinase 6) as a downstream transcriptional target based upon genetic deletion and pharmacologic inhibition studies. Treatment of these models with the dual CDK4/CDK6 inhibitor palbociclib resulted in inhibition of leukemia proliferation,⁹² although this therapeutic strategy has yet to be evaluated formally in the clinic.

A cryptic inversion of chromosome 16 (inv(16)(p13.3q24.3)) creates a fusion oncogene between the nuclear corepressor CBFA2T3 (CBFA2/RUNX1 partner transcriptional co-repressor 3, formerly ETO2) and the zinc finger DNA-binding transcription factor GLIS2 (GLI-similar family zinc finger 2) (Figure 1). CBFA2T3::GLIS2 is the most common fusion in the acute megakaryoblastic leukemia (AMKL) subtype occurring in pediatric patients without constitutional trisomy 21/Down syndrome and occurs nearly exclusively in infants and young children.^93,94 Preclinical studies have shown that the unique pediatric CBFA2T3::GLIS2 AMKL phenotype is determined by its developmental stage. Induction of CBFA2T3::GLIS2 in fetal HSCs results in AMKL development, whereas expression of the fusion oncogene in adult HSCs results in myeloid transformation with delayed leukemogenesis.⁹³ Additional studies using transduced cord blood have validated that CBFA2T3::GLIS2 fusions are sufficient to promote leukemogenesis, and co-occurring mutations are uncommon.^93,95 Clinically, CBFA2T3::GLIS2 AML is associated with marked chemoresistance and a dismal prognosis for children with 5-year OS <15%.²³ Recently, rare CBFA2T3::GLIS3 fusions have been reported in infants with AML or mixed phenotype acute leukemia that were also associated with resistance to multiple chemotherapy regimens and mortality due to progressive leukemia.^96,97 RNA sequencing of CBFA2T3::GLIS3 cases interestingly demonstrated similar gene expression profiles to those of patients with canonical CBFA2T3::GLIS2 fusions.⁹⁶ Identification of rare fusions in such cases highlights the need for unbiased diagnostic RNA sequencing of all childhood AML cases.

Very recent preclinical studies have elucidated a critical dependency of CBFA2T3::GLIS2 AMKL upon apoptosis pathways involving BCL2 and BCL-xL proteins and the therapeutic potential of their targeting with venetoclax and navitoclax, respectively.^95,98 Although preclinical leukemia model data suggest dual targeting of BCL2 and BCL-xL proteins with navitoclax may be more effective than treatment with venetoclax (which targets BCL2 alone),^95,98 a recent institutional case series remarkably reported CR in 3 of 4 (75%) children with relapsed/refractory CBFA2T3::GLIS2 AMKL treated with azacitidine, venetoclax, and GO that was second HSCT-enabling for some patients and resulted in long-term survival.⁹⁹ An additional case report has also noted success of an azacitidine/venetoclax-based approach for a patient with CBFA2T3::GLIS2 AMKL.¹⁰⁰ 

Report of the RAM immunophenotype of CBFA2T3::GLIS2 AMKL (characterized by bright CD56, dim-to-negative CD38 and CD45, and absent HLA-DR surface expression) has prompted preclinical investigation of antibody-based and cellular immunotherapies.¹⁰¹ A recent analysis of combined transcriptome data of cord blood transduced with CBFA2T3::GLIS2 and primary samples from pediatric patients with AML identified genes encoding plasma membrane proteins that were differentially expressed in CBFA2T3::GLIS2 AMKL, but silent in normal hematopoiesis.¹⁰² The folate receptor-α (FOLR1) was selected for further study given its expression in 80% of primary CBFA2T3::GLIS2 AMKL cases, as well as availability of FOLR1-targeted drugs with documented activity against solid tumors.¹⁰³ Folate receptors are plasma surface receptors that bind folate for endocytosis. Although FOLR2 and FOLR3 are expressed on granulocytes and monocytes, FOLR1 is enriched in cancers of epithelial origin with limited expression on normal tissues.¹⁰⁴ The specific mechanisms driving FOLR1 overexpression in CBFA2T3::GLIS2 remain incompletely elucidated at this time, but will be important to understand in light of new FOLR1-targeted therapies and potential mechanisms of treatment resistance that may ensue.

The FOLR1-directed antibody-drug conjugate (ADC) luveltamab tazevibulin (formerly STRO-002) has demonstrated safety and preliminary activity in a phase 1 trial in women with relapsed/refractory ovarian cancer¹⁰⁵ and was made available for children with relapsed/refractory CBFA2T3::GLIS2 AMKL via a compassionate use/extended access protocol given promising preclinical data.¹⁰⁶ In a recent report, 12 of 25 (48%) children treated on the extended access protocol with single-agent luveltamab or in combination with chemotherapy achieved MRD-negative CR, which was transplant-enabling in several patients.¹⁰⁷ The international multisite REFRaME-P1/COG ADVL2322 phase 1 trial is now evaluating the safety and preliminary activity of luveltamab in combination with chemotherapy in children with relapsed/refractory CBFA2T3::GLIS2 AMKL (NCT06679582). Based upon similar success in preclinical models,¹⁰⁸ FOLR1-targeted chimeric antigen receptor T-cell immunotherapy will also be evaluated in an institutional pediatric CBFA2T3::GLIS2 AMKL phase 1 trial (KG Tarlock, personal communication, January 2025).

Comprehensive next-generation sequencing efforts have defined the genomic landscape of childhood AML, highlighting its molecular heterogeneity and identifying patterns of mutual exclusivity vs cooperation in somatic leukemia-associated alterations. Oncogenic fusions can be sole drivers of leukemogenesis in childhood AML subtypes, particularly those occurring in younger patients. Conversely, comutations frequently occur in other AML subtypes, and their impact upon leukemia biology, prognosis, and treatment outcomes is still evolving.

As a first example, rat sarcoma virus (RAS) pathway mutations (eg, KRAS, NRAS, and PTPN11) occur in >50% of patients with KMT2A-rearranged AML.^25,109 Despite their subclonal nature with variant allele frequencies usually of ≤0.3 and potential for gain or loss at relapse,^1,94 RAS pathway mutations have recently been associated with significantly worse outcomes in childhood KMT2A-rearranged AML.¹⁰⁹ Preclinical studies have accordingly shown that treatment of KMT2A-rearranged/RAS pathway-mutant AML patient-derived xenograft models with dual menin and MEK inhibition was more effective than monotherapies, a strategy primed for clinical translation.¹⁰⁹ 

As a second example, FLT3-ITD occurs in 70% to 80% of NUP98::NSD1 AML cases,¹¹⁰ and WT1 mutations are also common.^110,111 Preclinical studies in murine models of NUP98-rearranged AML have shown that FLT3-ITD comutations accelerate leukemogenesis and promote more aggressive AML kinetics.^81,110,112,FLT3-ITD comutations (usually subclonal¹¹³) with NUP98::NSD1 fusions and/or WT1 mutations (likely an early event given similar frequency to that of NUP98 rearrangement itself¹¹³) confer a significantly worse outcome than NUP98::NSD1 rearrangement alone with high induction failure rates and poor long-term survival of patients.¹¹⁰ Small case series of patients with relapsed/refractory NUP98–rearranged AML have reported absence of initially co-occurring FLT3-ITD at relapse (usually following targeted FLT3 therapy). The addition of sorafenib to chemotherapy in the COG AAML1031 clinical trial did not increase EFS or OS for children with NUP98-rearranged/FLT3-mutant AML compared to chemotherapy alone, which was in contrast to the improved clinical outcomes of patients with FLT3-ITD without NUP98-rearrangements.¹⁴ Additional preclinical studies have also reported potential therapeutic activity of navitoclax- and dasatinib-based treatment of NUP98-rearranged/FLT3-mutant AML,⁹⁴ although further studies are needed.

As a third example, RB mutations frequently occur in NUP98::KDM5A AML, which often presents as an AMKL morphology and immunophenotype.²³ Structural chromosome 13 abnormalities, including del(13q), monosomy 13, and translocations involving chromosome 13, have interestingly been associated with improved outcomes in patients with NUP98::KDM5A AML for currently incompletely understood reasons.²² This study also identified a strong association of differential survival of patients with NUP98::KDM5A AML depending upon the specific NUP98 break point; those with exon 13 break points had an OS of 51% vs 0% for exon 14.²² 

The examples above provide interesting observations on how co-occurring mutations can impact the prognosis of children with high-risk AML, although current risk stratification algorithms do not (yet) intercalate such data. Comprehensive next-generation sequencing analyses coupled with robust clinical outcomes and detailed preclinical mechanistic data will increase confidence in the ability to allocate patients to appropriate targeted therapies, ideally in the front line. Further advances in modern genomic technologies, including whole genome sequencing and single-cell RNA and DNA sequencing assays that can more clearly define leukemia cell heterogeneity and determine if fusions and/or mutations are present in primitive AML cell driver populations vs more differentiated blasts, will be further illuminating.

Exciting progress has been made in the past decade in advancing biologic and clinical knowledge of oncogenic fusion-driven childhood AML and ascertaining how genetic alterations can be matched with appropriate precision medicine therapeutics. The emerging success of FLT3 inhibitors, menin inhibitors, BCL-2 inhibitors, and CD33- or FOLR1-targeted immunotherapies provides proof-of-principle that detailed genomic and functional characterization of critical leukemia biology coupled with understanding of associated clinical outcomes is essential for modern risk stratification and targeted therapy success. Although increasingly smaller “boutique” childhood AML subtypes have been identified through such efforts, critical shared pathways and principles persist nonetheless. Ideally, successful development of biologically rational therapeutic drug combinations targeting oncofusion-driven aberrant pathway dysregulation and essential co-occurring mutational dependencies, will decrease relapse risk and improve long-term cures of children with high-risk genetic subtypes of AML.

G.E. is supported by an Ontario Institute of Cancer Research (OICR) Investigator Award. This study was conducted with the support of OICR through funding provided by the Government of Ontario. G.E. is also supported by the Leukemia and Lymphoma Society of Canada, Kindred Foundation, and the Garron Family Cancer Centre. S.K.T. is supported by the National Institutes of Health/National Cancer Institute awards (1U01CA232486 and 1U01CA243072), United States Department of Defense Translational Team Science Award (CA180683P1), a Pennsylvania Department of Health Commonwealth Universal Research Enhancement (CURE) award, the Leukemia & Lymphoma Society, and the Andrew McDonough B+ Foundation. S.K.T. is a Scholar of the Leukemia &Lymphoma Society and holds the Joshua Kahan Endowed Chair in Pediatric Leukemia Research at the Children’s Hospital of Philadelphia.

The views expressed in the publication are the views of the authors and do not necessarily reflect those of the Government of Ontario.

Contribution: G.E. and S.K.T. conceived and directed the study, reviewed scientific literature, and wrote and edited the manuscript; and both authors reviewed and approved the manuscript before submission.

Conflict-of-interest disclosure: S.K.T. receives research funding from Kura Oncology for preclinical investigation of ziftomenib in pediatric leukemias; has served on scientific advisory boards for Kura Oncology for clinical investigation of ziftomenib and Syndax Pharmaceuticals for clinical investigation of revumenib; and is an institutional principal investigator at the Children’s Hospital of Philadelphia for early-phase clinical trials of revumenib, ziftomenib, and luveltamab tazevibulin for children with relapsed/refractory acute leukemias. G.E. declares no competing financial interests.

Correspondence: Grace Egan, Division of Haematology/Oncology, The Hospital for Sick Children and University of Toronto, 175 Elizabeth St, 10-14-066, Toronto, ON M5G 2G3, Canada; email: grace.egan@sickkids.ca; and Sarah K. Tasian, Children's Hospital of Philadelphia, Division of Oncology & Center for Childhood Cancer Research, 3501 Civic Center Boulevard, CTRB 3056, Philadelphia, PA 19104; email: tasians@chop.edu/ s.k.tasian@prinsesmaximacentrum.nl.