Current perspectives on KMT2A fusion proteins and menin inhibition in paediatric acute myeloid leukaemia.

Introduction
Acute myeloid leukaemia (AML) is a molecularly heterogeneous cancer of the bone marrow and peripheral blood that is characterised by the (oligo−/multi‐)clonal proliferation of undifferentiated myeloid blasts or progenitor cells [1, 2]. It tends to be viewed as a disease of the elderly; however, it can develop in children, sometimes as early as a few days after birth [3] or even in utero, as translocations have been observed on Newborn Screen/Guthrie test blood spots [4]. Despite some advances in therapeutic options, AML remains difficult to treat. This is particularly the case in very young patients with increased susceptibility to therapy‐related toxicity [5]. Fortunately, collaborative efforts are already in place to better understand the biology of several high‐risk or hard‐to‐treat paediatric cancers. One of which is the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative. A landmark study from their AML project involved genomic analyses of nearly 1000 children and young adults with AML and revealed that the disease differs greatly between younger and older patients, at the genetic level. Specifically, somatic mutations are much more common in older adults with AML, whereas structural alterations are more abundant in younger patients with the disease [6]. A substantial proportion of the structural alterations are due to translocations, where sections of chromosomes shift, and can result in the production of fusion genes and fusion proteins. Indeed, rearrangements involving chromosome band 11q23 and KMT2A (KMT2Ar) are key drivers of acute leukaemias—representing about 15% of paediatric AML cases [7]—and a defining genetic abnormality used by the World Health Organization to help classify AML [8]. These patients often present with high‐risk clinical characteristics, including elevated white blood cell counts and extramedullary involvement [9].
NB: Human Genome Organisation (HUGO) nomenclature (https://www.genenames.org/) is utilised throughout the text, with some genes along with their former synonyms listed here for convenience: KMT2A = former MLL, AFF1 = former AF4, MLLT1 = former ENL, MLLT3 = former AF9, AFDN = former AF6, MLLT10 = former AF10, MLLT11 = former AF1Q.

KMT2A structure and function

KMT2A, or MLL1 as it was previously known, encodes a histone H3 lysine 4 methyltransferase which plays a crucial role in regulating gene expression during embryonal development and the maintenance of haematopoietic stem cells [10]. The KMT2A protein is processed by the endopeptidase, Taspase1, into two amino (N320) and carboxyl (C180) terminal fragments [11], which reassociate (via their FYRN and FYRC domains) to form a molecular hub for the assembly of larger multiprotein complexes in the nucleus (Fig. 1). Simply put, the N‐terminal portion of KMT2A is responsible for promoter binding and reading of chromatin signatures, while the C‐terminal portion—which contains a Su(var)3–9, Enhancer‐of‐zeste and Trithorax (SET) domain—methylates lysine 4 within histone H3 to promote the transcription of target genes, including HOX genes [12, 13] (Fig. 1). It is worth noting that transcriptional profiling of isogenic wildtype and KMT2A knockout murine fibroblasts revealed that more genes were upregulated than downregulated in the knockout situation, suggesting the presence of a molecular switch that toggles between transcriptional activation and repression [14]. We now know that this is due to a conformational change following CYP33—a prolyl‐peptidyl isomerase—binding to the enhanced plant homeodomain 3 (ePHD3) subdomain within KMT2A. Typically, ePHD3 binds to di‐ and tri‐methylated lysine‐4 residues within histone H3 [15]. However, CYP33‐mediated cis‐trans isomerisation of the proline‐1665 residue in KMT2A prevents this and instead enables the subsequent recruitment of a polycomb repressor complex (i.e. BMI1, HPC2, CtBP and HDAC1/2) to the methyl‐DNA binding domain (MBD) of KMT2A [16] (Fig. 2). The distinct domains within KMT2A which we are still learning more about [17], along with the various proteins and transcription factors that contribute to the functional KMT2A multiprotein complex, are crucial for regulating target genes during embryonic haematopoietic stem cell development [18] and adult haematopoiesis [10].
Given the subtle but fundamental roles of wildtype KMT2A in haematopoiesis, it is no surprise that genetic rearrangements resulting in the expression of KMT2A fusion alleles can lead to dramatic transcriptional disturbances that contribute to the onset of acute leukaemias. Indeed, aberrant H3K79 methylation profiles [19] and upregulated HOXA/MEIS1 gene signatures [20] are hallmark features and well‐established pathomechanisms of KMT2Ar acute leukaemias. While KMT2A rearrangements affect the vast majority (≥ 70%) of infant acute lymphoblastic leukaemia (ALL) cases and a substantial proportion of adult cases, KMT2Ar ALL is beyond the scope of this article, and interested readers are instead directed to [21, 22, 23] for further information.
Chromosomal translocations involving human KMT2A (NM_001412597.1) and 112 partner genes have been identified so far [17]. However, the majority of recurrent KMT2Ar leukaemic translocations involve relatively few fusion partner genes. For instance, ~ 80.7% of cases in a paediatric cohort of 1256 patients with KMT2Ar AML were characterised by six fusion genes, including KMT2A::MLLT3/t(9;11) (p22; q23) (43.3%), KMT2A::MLLT10/t(10;11) (p12; q23) (17.4%), KMT2A::AFDN/t(6;11) (q27; q23) (7.3%), KMT2A::ELL/t(11; 19) (q23; p13.1) (6.0%), KMT2A::MLLT1/t(11; 19) (q23; p13.3) (4.5%) and KMT2A::MLLT11 t(1; 11) (q21; q23) (2.2%) [24]. Interestingly, four of these partner proteins (i.e. MLLT3/AF9, MLLT10/AF10, ELL and MLLT1/ENL) are involved in transcriptional elongation, as they either bind directly or indirectly—via Super Elongation Complexes (SEC)—to RNA polymerase II [25, 26, 27, 28].
With such a comprehensive list of KMT2A fusion partners, two key concepts emerge—what are the molecular consequences of KMT2A translocations and how important are the contributions of the different fusion partners to the underlying cancer biology of KMT2Ar AML?
With regards to the first point, KMT2A rearrangements usually result in physical separation of the MBD from downstream plant homeodomain (PHD) regions, resulting in loss of the intrinsic control mechanism of the KMT2A protein and separating its epigenetic reading and writing functions. Consequently, both separated portions of KMT2A become constitutively active, regardless of their fused protein sequences. The direct KMT2A‐X fusions still bind via menin/LEDGF and the PAF complex to chromatin but are incapable of transcriptional repression as they lack the necessary PHDs (Fig. 2). This was effectively demonstrated by the artificial fusion of PHD to existing KMT2A‐X proteins, which then enabled the subsequent recruitment of the polycomb repressor complex and eliminated the oncogenic properties of the KMT2A‐X fusion proteins [29, 30]. Reciprocal X‐KMT2A fusion proteins which retain their ePHD3 chromatin reader and SET domain are unable to bind the polycomb repressor complex, regardless of CYP33 binding, due to loss of the MBD (Fig. 2). This is not dissimilar to other chromosomal translocations, such as the oncogenic BCR‐ABL1 translocation in chronic myeloid leukaemia which destroys an intrinsic control mechanism within the ABL kinase, producing a constitutively active BCR‐ABL1 fusion protein [31].
Delving beyond the physical separation of the MBD and PHD, the KMT2A gene contains a major breakpoint cluster region (BCR1) between intron 7 and exon 13, and a minor BCR (BCR2) located further downstream between exons 21 and 25. Interestingly, while breakpoints in ALL patients are found in both regions, AML breakpoints seem to exclusively occur in BCR1 [32]. The focus on KMT2A breakpoint distribution stems from correlations with the outcome of KMT2Ar leukaemia patients. More specifically, breakpoints within exon 11 or intron 11 of KMT2A are typically associated with poorer clinical outcomes [33]. This is because these breakpoints can alter or destroy the structure of the cysteine‐histidine‐rich PHD1‐3 domain thereby compromising its associated functions, which can: (a) limit functional dimerisation [34]; (b) enhance the stability of KMT2A fusion proteins by interfering with the binding of CDC34 (to PHD2) [35] or ECSASB2 [36], thereby rendering them resistant to proteasomal degradation; and (c) impair CYP33 binding leading to loss of the transcriptional repressor activity of KMT2A [37]. While AML patient breakpoints exhibit a preferential KMT2A intron 9 breakage which decreases slightly with age, further analysis of breakpoint localisation within subregions of BCR1 (i.e. exon9‐intron10 and exon11‐exon13), revealed that certain fusion genes have selective breakpoint preferences (which probably has something to do with the respective functions of the fusion partners) and that these breakpoint tendencies also seemed to change with age [32]. For example, KMT2A::AFF1 and KMT2A::ELL patients exhibit opposite predispositions, that is infant KMT2A::AFF1 patient breakpoints predominantly localise to intron 11, while adult KMT2A::AFF1 patient breakpoints shift to introns 9 and 10; whereas infant KMT2A::ELL patients have a preference for intron 9 breakpoints, while paediatric and adult KMT2A::ELL patient breakpoints shift towards intron 11 [32].
Analysis of large cohorts of patients with integrated molecular data suggests that KMT2A fusion partners carry prognostic significance. For instance, KMT2A::MLLT3 and KMT2A::MLLT11 tend to be associated with more favourable survival outcomes in children, whereas KMT2A::MLLT10 and KMT2A::AFDN are among the KMT2A rearrangements with the poorest prognoses in children [38]. Furthermore, KMT2A::MLLT3 is categorised as intermediate risk according to the 2022 edition of the European Leukaemia Net recommendations, with other KMT2Ar associated with less favourable outlooks [39]. Although it should be noted that the prognosis of children with KMT2A::MLLT3 can be adversely affected by the presence of secondary chromosome aberrations [24, 40] and/or by a marrow morphology that is different from acute monoblastic leukaemia (FAB‐M5) [38]. In a similar vein, more recent evidence suggests that the aforementioned favourable risk stratification of KMT2A::MLLT11 should be revised to intermediate risk [24]. A retrospective international study of 1130 paediatric KMT2Ar AML patients categorised cases with KMT2A::MLLT10, KMT2A::AFDN, KMT2A::MLLT1, KMT2A::AFF1/t(4; 11) (q21; q23), and KMT2A::ABI1/t(10; 11) (p11.2; q23) rearrangements as high risk, with remaining cases allocated to a non‐high risk group [9]. The high‐risk cohort exhibited a higher cumulative incidence of relapse (59.7% for high risk vs 35.2% for non‐high risk; P < 0.0001), lower event‐free survival (EFS) rates (30.3% vs 54.0%; P < 0.0001), and worse overall survival (OS) (49.2% vs 70.5%; P < 0.0001) [9].
The remaining list of recurrently diagnosed translocation partner genes is long, but these cases are rare and not of pressing clinical relevance. However, they should not be overlooked and a systemic classification about their functions has already been proposed [41]. The presence of reciprocal X‐KMT2A fusion proteins is yet another important factor, as demonstrated by the chromatin opening capability of AFDN::KMT2A and its ~ 12‐fold enhancement of the KMT2Ar transcriptome [42]. Gene dysregulation of this magnitude likely contributes to the onset of pre‐leukaemic clones that then undergo selection to overt leukaemic cells. However, the enhanced transcriptome could also result in the expression of more druggable target proteins that may sensitise the leukaemic cells, ultimately translating to better clinical outcomes. It is possible that leukaemic cells lacking a reciprocal fusion protein and therefore possessing a more restricted transcriptome, might be more chemo‐resistant. As such, the identification of reciprocal fusion proteins could provide important prognostic information and impact disease monitoring.

Current and future therapies involving menin inhibition
Children with KMT2Ar AML are associated with an intermediate 5‐year event‐free survival rate of ~ 45% and an overall survival rate of ~ 63%, highlighting an unmet clinical need for this patient population [9]. At present, strategies to treat KMT2Ar leukaemias are based on canonical risk stratification [39], although menin inhibition has emerged as a promising new class of targeted therapy for KMT2Ar AML [2].

Biological basis for menin inhibition
Menin is a chromatin adaptor protein, encoded by the MEN1 gene, that participates in epigenetic gene regulation through its interactions with various proteins—including chromatin modifying proteins and transcription factors [43]. It has paradoxical roles, acting as both a tumour suppressor and an oncogenic co‐factor, depending on the cellular context. Mutations in the MEN1 gene that disrupt its functionality are typically associated with Multiple Endocrine Neoplasia Type 1 (MEN1) syndrome [44, 45], highlighting the tumour suppressive activity of the gene in neuroendocrine tissues. In contrast, menin serves as an oncogenic co‐factor in haematopoietic pathological states [46]. Indeed, menin was initially identified as an essential oncogenic co‐factor for KMT2Ar leukemogenesis in a mouse model back in 2005 [47]. Since then, there has been considerable development into several iterations of highly selective small molecule menin inhibitors (summarised in Table 1), with eight compounds currently at various clinical stages for the treatment of AML: ziftomenib (KO‐539), revumenib (SNDX‐5613), bleximenib (JNJ‐75276617), BN104, icovamenib (BMF‐219), DS‐1549, enzomenib (DSP‐5336) and balamenib (ZE63‐0302) (summarised in Table 2). The MEN1 inhibitor (or MI) class of drugs (listed in Table 1) has culminated in the commercially available ziftomenib, whereas the preclinical precursor VTP‐50469 has led to revumenib which is currently the only FDA‐approved menin inhibitor for the treatment of relapsed/refractory (R/R) AML [48]. With a generally favourable safety profile in patients, most of the adverse effects associated with these agents are manageable and reversible through dose adjustments and supportive care [2].
These compounds work by specifically targeting a central hydrophobic pocket on menin's surface to prevent the protein–protein interaction between menin and KMT2A (i.e., either the N‐terminal portion of wildtype KMT2A or oncogenic fused versions of the protein) [49]. This effectively inhibits KMT2A‐dependent transcription of downstream target genes, thereby blocking an important component in the pathogenesis of KMT2Ar AML. Importantly, the first two menin inhibitors taken forward for in vivo work—MI‐463 and MI‐503—blocked the progression of KMT2Ar AML in mice, without impairing normal haematopoiesis [50]. Furthermore, another study found that treatment of KMT2Ar AML cell lines with the same two menin inhibitors led to increased ubiquitylation and reduced protein stability of menin, suggesting an additional aspect of menin inhibition beyond physical disruption of the menin‐KMT2A interaction [51]. Leaning into this theme, recent work has demonstrated altered gene expression following menin inhibition, beyond the typical silencing of the HOX‐ and MEIS1‐dependent oncogenic transcriptional programme. For instance, menin loss (via genetic knockout and VTP‐50469 treatment) resulted in redistribution of KMT2A from (menin‐dependent) active genes to a subset of silent bivalent genes (i.e. genes which are concurrently marked by activating H3K4me3 and repressive H3K27me3 modifications to maintain a transcriptionally inert state, poised for activation or stable repression), which elevated menin‐independent KMT2A gene expression [52]. Moreover, these upregulated genes included major histocompatibility class I (MHC‐I) molecules, suggesting that the efficacy of menin inhibition may involve the loss of immune evasion in KMT2Ar leukaemic blasts [52]. Additional supporting evidence comes from a recent study demonstrating bleximenib‐induced (MEIS1‐independent) upregulation of HLA class I and II expression in leukaemic blasts, along with enhanced T‐cell mediated cytotoxicity [53].
In summary, the anti‐leukaemic effects of menin inhibitors can be achieved through both canonical and non‐canonical mechanisms, via disruption of leukaemogenic transcriptional programs while also reshaping the immunogenic landscape of KMT2Ar AML. As such, menin inhibition remains a promising therapeutic strategy with potential to directly suppress leukaemic proliferation and enhance immune‐mediated clearance of leukaemic cells.

Challenges and resistance mechanisms
Menin inhibitors have great potential for addressing unmet needs in various subtypes of genetically defined acute leukaemias, typically characterised by (menin‐dependent) upregulation of HOX gene expression [54]. However, enrolling paediatric patients in clinical trials is challenging, mainly due to the small number of affected individuals. The annual incidence of newly diagnosed paediatric AML cases globally is fewer than 1000, with an estimated annual incidence of around 350 cases of newly diagnosed and recurrent KMT2Ar AML among paediatric patients [55]. Without global collaboration, these limitations can impede insights into side effects and efficacy across KMT2A fusion partners.
Drug resistance is a recurring theme in the development of cancer treatments, and menin inhibition is no exception. Two forms of menin inhibitor resistance have been described: MEN1 mutations and non‐MEN1 mutation‐driven mechanisms for example through cellular adaptation processes. In the case of the former, somatic MEN1 mutations were identified in patients that received revumenib (38.7% in the AUGMENT‐101 phase 1 trial), as well as in patient‐derived xenograft models and an unbiased base‐editor screen [56]. Some of these mutations (including M327I, G331D, G331R and T349M) affect residues near the KMT2A binding site in menin, generating steric hindrance with revumenib to effectively attenuate binding of the inhibitor with menin, but without impacting the menin‐KMT2A interaction [57]. The G331D mutation is particularly interesting as it results in a very slow dissociation of KMT2A from menin, suggesting that the meninG133D‐KMT2A interaction might be even more challenging to dissociate with small molecule inhibitors, when compared to wildtype and other mutant versions of menin [57].
Menin inhibitor resistance arising in the absence of MEN1 mutations is much broader, ranging from alterations in parallel or downstream pathways [58, 59] to treatment‐induced clonal evolution despite on‐target efficacy of inhibitors at the transcriptional level [60, 61]. Delving deeper into the altered signalling pathways, loss of non‐canonical polycomb repressive complex 1.1 (PRC1.1)‐mediated signalling has been linked with menin inhibitor resistance, via epigenetic reactivation of non‐canonical menin‐KMT2A targets, such as MYC [58]. This is problematic because MYC is a critical oncogene that has been implicated in AML pathogenesis [62] and the promotion of cytotoxic drug resistance [63]. Moreover, after establishing that the UTX‐KMT2C/D complex contributes to the efficacy of menin inhibition by inducing a myeloid differentiation program, a patient‐derived xenograft model of AML harbouring a KMT2C mutation was found to be resistant to VTP‐50469 [59]. This effectively highlights acquired mutagenesis of essential, non‐driver epigenetic regulators as potential adaptive pathways contributing towards menin inhibitor resistance.

Strategies to mitigate resistance
Some of the MEN1 mutations (i.e. M327I and T349M) that confer resistance to revumenib could potentially be addressed with newer iterations of menin inhibitors, such as bleximenib (JNJ‐75276617) [64] and balamenib (ZE63‐0302) [65] which circumvent the acquired steric hindrance through a unique binding mode. With this in mind, it seems prudent to standardise MEN1 mutation status monitoring during menin inhibition therapy, as this could help guide treatment decisions and ultimately improve patient outcomes.
Combination treatments with menin inhibitors are also being pursued both in preclinical studies (Table 3) and clinical trials (Table 2) to either prevent or overcome intrinsic and acquired mechanisms of resistance.
While KMT2Ar AML typically has a lower incidence of co‐existing gene mutations, a number of tyrosine kinase‐PI3K‐RAS pathway mutations (FLT3, NRAS, KRAS, PTPN11 and BRAF) have been reported to co‐occur and confer more aggressive, high‐risk leukaemic profiles [66, 67], presumably by promoting the clonal expansion of KMT2Ar leukaemic cells, as demonstrated in a retroviral AML mouse model [68]. Although resistance to menin inhibitors in these unfavourable AML subsets hasn't yet been described per se, combination therapies with menin and FLT3 or MEK inhibitors (which target the RAS/MAPK pathway) are already underway in clinical (Table 2) and preclinical [69] (Table 3) settings.
The rationale for combining menin inhibition with broadly effective agents (e.g., chemotherapy, azacitidine, or venetoclax) is based on leveraging distinct mechanisms of action to achieve both cytotoxic debulking and differentiation of leukaemic cells. Interestingly, these combination treatments may even help mitigate the risk of differentiation syndrome [70], which is a commonly occurring drug‐related adverse event associated with menin inhibitor monotherapies (Table 2). While there is a logical basis for combining menin inhibition with current standard‐of‐care approaches, additional preclinical work is required to better understand the mechanistic basis of and optimise any additive/synergistic effects. Further research should explore the optimal sequencing of these double, triple and even quadruple combination treatments (including other targeted therapies i.e. FLT3 and MEK inhibitors): for instance, whether the agents should be given concurrently or sequentially, and whether certain components could be dropped over time. Comparative evaluation of differentiation syndrome rates, myelosuppression, QTc impact, minimum residual disease clearance and response durability will be crucial going forward. These evolving strategies should continue to expand our understanding of other resistance pathways and hopefully pave the way for novel or improved therapeutic approaches to refine treatment paradigms and enhance remission rates.

Clinical positioning of menin inhibitors
Menin inhibition is one of the latest additions to the growing arsenal of targeted AML treatments, particularly for subsets of patients harbouring KMT2A rearrangements or NPM1 mutations. The precise mechanism underpinning the reliance of this latter AML subtype on the menin‐KMT2A interaction is unclear but is presumably due to a loss of nuclear function (as mutated NPM1 persists in the cytoplasm), culminating in upregulation of HOX gene expression which in turn drives leukemogenesis [71, 72]. Regardless, the regulatory pathway for KMT2A and NPM1‐mutated AML subtypes is more clearly defined and is probably why commercial development efforts have been focused on these patient cohorts. However, there is a broader spectrum of patients (e.g. those with NUP98 or NUP214 rearrangements, UBTF tandem duplications and other HOX‐mediated leukaemias) who could benefit from menin inhibition as the field evolves, due to the dependency of their leukaemic cells on the menin‐KMT2A interaction.
Menin inhibitors are currently undergoing investigation as monotherapies or in combination with intensive chemotherapies (typically for younger, fit patients) and more targeted treatment regimens (for older and/or unfit patients). Clinical exploration has centred on menin inhibition within the R/R AML setting (Table 2), and this is further underscored by the FDA's approval of revumenib for KMT2Ar R/R AML in patients aged 1 year or older [48]. However, there is growing interest in expanding into frontline regimens for newly diagnosed AML and potentially utilising menin inhibitors as maintenance therapy, even after allogeneic stem cell transplantation, where appropriate. Additionally, the effectiveness of menin inhibitors in paediatric patients is also being explored via the extension of adult study criteria to include younger individuals in clinical trials (bolded trials in Table 2). This will hopefully accelerate the establishment of safe and effective paediatric dosages, potentially in combination with appropriate multi‐agent chemotherapy. Collectively, the data from early‐stage clinical trials is promising and demonstrates proof‐of‐concept; however, it is clear that larger, randomised Phase III studies are necessary for validation. Furthermore, there remains a lack of long‐term data on the durability of clinical responses, the potential for late‐emerging toxicities and overall survival benefits [73].

Outlook and future considerations
The expansion of targeted treatment options for AML offers improved efficacy and fewer toxicities, culminating in a greater quality of life for patients. Menin inhibition will undoubtedly reshape the treatment paradigm for patients with KMT2Ar and other menin‐sensitive genetic alterations that were beyond the scope of this review, including: NPM1 mutations, NUP98 and NUP214 rearrangements, and UBTF tandem duplications. To help achieve this, the creation of a predictive/pharmacodynamic biomarker assay that is capable of reliably detecting HOX gene expression (or a characteristic gene signature based on menin‐KMT2A binding motifs) would be ideal for: (a) identifying and expanding into a larger subset of patients that may benefit from menin inhibition; (b) predicting the responses of a patient to different menin inhibitors ex vivo to better inform and tailor treatment regimens; and (c) helping to monitor treatment responses (which should also be supplemented with standardised prospective monitoring of MEN1 mutation status). Furthermore, while the emergence of drug resistance is inevitable, ongoing research into combining menin inhibitors with other targeted therapies that could prevent and/or overcome acquired resistance is encouraging. However, realising the full potential of menin inhibition within the paediatric/childhood setting of AML will require global co‐operation. Securing ethical approval for trial initiation, overcoming patient recruitment and enrolment hurdles, sharing data and resources, as well as forging collaborative partnerships between industry, academic institutions, government agencies and clinicians are all vital components for accelerating progress and ultimately ensuring broad access to this promising therapy for all children and adolescents with KMT2Ar AML.

Conflicts of interest
LER and KML have no conflicts of interest to declare. GG received an honorarium from Novartis, GSK. GG also reports financial support in the form of an Education Support/Travel award from Novartis.

Author contributions
Drafting of the manuscript and articulation of the primary content were undertaken by LER. Revisions to enhance the intellectual content and ensure accuracy were performed by LER, GG and KML. All authors have reviewed and approved the manuscript in its final form.