KMT2A-rearranged leukemia: from mechanism to drug development

Abstract

Gene rearrangements of the human MLL gene (also known as KMT2A) generate multiple fusion oncoproteins which cause leukemia with poor prognosis. MLL is an epigenetic regulator that reads and writes epigenetic information and has an evolutionarily conserved role maintaining expression of Homeotic (HOX) genes during embryonic development. Most MLL gene rearrangements found in leukemia generate a constitutively active version of the wild-type protein, which causes overexpression of HOX and other genes and leukemic transformation of normal hematopoietic progenitors. Elucidating the molecular mechanisms underlying how MLL activates gene expression and how gene rearrangements affect this gene-regulating activity provided therapeutic opportunities to block fusion oncoprotein-specific activities. One uniform molecular dependency of MLL fusion oncoproteins is its interaction with the chromatin-binding partner MENIN that is essential to maintain leukemic transformation. MENIN inhibitors that interfere with the MLL-MENIN interaction have been developed and are now entering clinical practice. Also, the MLL complex physically interacts with several histone acetyl transferases (HATs), including MOZ/MORF, HBO1, and EP300/CREBBP to effect MLL-MENIN-dependent gene activation. Aberrant recruitment of these HATs and other transcriptional effector complexes are key differences between MLL and MLL fusion oncoproteins. In this review, we first summarize our current understanding of wild-type MLL function and aberrant function of its oncogenic variants. We then discuss in detail how chromosomal translocations generate constitutive active forms of MLL and categorize them into five major classes. We touch on the collaborative gene activation by MLL and specific interacting HATs. Lastly, we discuss how these mechanistic insights have led to the development of the first-in-class MENIN inhibitors and discuss efforts to anticipate and treat both genetic and non-genetic mechanisms of resistance.

Graphical Abstract

Teaser abstract

MLL fusion oncoproteins often transform cells in particular developmental states without the need for additional mutations. Insights gained studying how this occurs has yielded a wealth of mechanistic and biological insights. Here we review some of the detailed molecular mechanisms leading to leukemia development as well as how understanding these mechanisms resulted in the development of targeted therapies, and ongoing efforts to anticipate and prevent resistance to targeted therapies

Introduction

The human MLL gene (also known as KMT2A) drew attention due to its discovery at the site of recurrent chromosomal translocations in childhood leukemia [1–3]. The first clues to its biological function arose from its similarity to the Drosophila trithorax gene, which had been studied for its role in heritable gene expression patterns for over 40 years [4]. The similarity suggested that MLL and trithorax share common functions, which was supported generally by loss-of-function studies in mice and human cells [5–9]. Trithorax maintains expression of Homeotic (Hox) genes during embryo development to confer identity of body segments, a feature shared with MLL [5]. MLL and trithorax recognize a unique combination of DNA, epigenetic marks and protein interactions. In the early 2000s, it was appreciated that trithorax and MLL encode histone methyltransferases via a conserved Su(var)3–9/Enhancer of Zeste/Trithorax (SET) domain [10, 11]. It is now understood that these proteins maintain gene expression through catalytic and non-catalytic mechanisms.

In acute leukemia, chromosomal translocations generate MLL fusion proteins in which the MLL-N terminus is fused in-frame to one of many partners. Numerous translocations and gene rearrangements have been described, although the most common t(4;11), t(9;11) and t(11;19) account for the majority of cases [12]. An additional MLL gene rearrangement, termed MLL-PTD for Partial Tandem Duplication, is also transforming despite the absence of a fusion partner [13, 14]. Due to the variety of C-terminal fusion partners and the transforming activity of the MLL-PTD molecule, significant effort has gone into understanding how the wild-type (WT) MLL functions normally in hematopoietic cells and how its functions are altered through the various gene rearrangements that lead to oncogenic activity.

Specifically, investigators have asked: 1) are there common biochemical features of MLL fusion oncoproteins that would support a unified theory explaining mechanism(s) of transformation?, 2) how are the oncoprotein-driven transcriptional pathways distinct from those sustained by the proto-oncogene? and 3) what differences between oncogene and proto-oncogene can be exploited to develop the most specific, molecularly-targeted anti-leukemia therapy?

In this review, we first summarize current understanding of the normal biological and biochemical functions of the MLL proto-oncogene. We then discuss how chromosomal translocations deregulate these functions through MLL fusion oncoproteins, categorizing the major fusion classes based on shared molecular features and mechanisms of transcriptional activation. Finally, we discuss how these mechanistic insights have informed the development of targeted therapies, culminating in the clinical approval of the first-in-class MENIN inhibitors and emerging strategies to overcome resistance.

Normal biological and biochemical roles of the MLL proto-oncoprotein

MLL is a large nuclear protein predicted to be 430 kDa. The MLL protein is cleaved by a cellular protease, Taspase I, shortly after translation, resulting in two major fragments: MLL^N and MLL^C [15–17]. These two fragments form an intra-molecular complex through the structures of the FYRN-, FYRC-domains, and PHD finger 1 and 4 [16, 18]. A genetically engineered mouse line carrying mutations preventing this cleavage develops normally with no obvious phenotypes [19, 20], suggesting a subtle or tissue specific role for this processing event. The extreme C-terminus harbors the conserved, enzymatically active SET domain that nucleates an enzymatically active histone methyltransferase complex [21, 22] whereas the N terminus harbors multiple DNA/chromatin interacting motifs (Figure 1).

Figure 1. Structures and functions of MLL related proteins.

A) WT MLL protein diagram compared to oncogenic forms and the Drosophila homolog, trithorax. Arrows and dotted lines indicate recruited proteins and their corresponding enzymatic activity on chromatin. Light green spheres indicate non-constitutive associating factors that bind in a sub-stoichiometric manner. B) Diagram of WT MLL protein assembly with protein partners and association with chromatin at a model promoter. Menin is highlighted in red.

The SET domain of WT MLL binds ASH2L, RbBP5, and WDR5, which are required for the mono-, di- and tri-methylation of Histone H3 lysine 4 (H3K4) [10, 23, 24] (Figure 1). It was anticipated that this MLL-mediated histone methylation would play a significant role in gene regulation. However, a genetically engineered mouse lacking the SET domain and hence enzymatic activity showed only a modest reduction of Hox gene expression and minor skeletal defects, surviving to adulthood with normal hematopoiesis [25, 26], contrasting to the severe embryonic lethal phenotypes of MLL knockout mouse strains [5, 18, 19, 27]. This indicates that SET domain-mediated H3K4 methylation is not a major contributor to the regulation of target genes, at least in the hematopoietic system. MLL also possesses a potent transcriptional activation domain in the MLL^C portion, which binds to CREBBP/EP300 histone acetyltransferases [28]. CREBBP/EP300 acetylates Histone H3 lysine 18/27 (H3K18/27ac) to activate transcription [29]. In addition, this C-terminal portion has been reported to interact with MOF, and other potential effectors of transcription [30]. A mouse line lacking the entire MLL^C fragment cannot express Hox genes during embryogenesis and dies in utero [18], indicating that the functions of the C-terminal fragment are critical. The processing sites, intra-molecular interaction domains including FYRN, FYRC, and PHD fingers, SET domain, and transcriptional activation domain are evolutionally conserved between human and Drosophila, suggesting fundamental biological roles for these interactions [9] (Figure 1A).

MLL has many domains that target to different chromatin structures. One of the chromatin-targeting activities of MLL occurs through its interaction with MENIN [24, 31], which leads to further association with LEDGF (also known as PSIP1) [32]. LEDGF has a PWWP domain, which is a chromatin reader module that recognizes di-/tri-methylated Histone H3 lysine 36 (H3K36me2/3) [33, 34] (Figure 1A). H3K36me2/3 marks are indicative of transcriptionally active chromatin [35]. Hence, LEDGF recruits the MLL/MENIN complex to transcriptionally active promoters through this specificity. MLL also harbors a CXXC domain, which directly binds to unmethylated CpGs [36–39]. CpG islands are associated with a majority of mRNA promoters and facilitate the establishment of bivalent chromatin, but can also be subject to methylation-mediated repression mechanisms [40] Chromatin harboring unmethylated CpGs and H3K36me2/3 defines a core requirement for the N-terminus to target chromatin. In addition, most genome-wide localization studies find that MLL occupancy overlaps significantly with transcriptionally active marks and is enriched at transcription start sites (TSS) [41, 42]. A plant homeodomain (PHD) finger in MLL specifically binds to H3K4me3, likely further stabilizing the interaction of the MLL complex to promoters [23, 36, 43]. Consequently, the genomic distribution of MLL spans many transcriptionally active promoters [42]. To a first approximation, the WT MLL complex binds to transcriptionally active promoters containing unmethylated CpGs, H3K36me2, and H3K4me3 enriched chromatin. This localization is quite non-specific, whereas a much smaller subset of genes critically depend on MLL for their expression, suggesting additional layers of regulation through cofactor recruitment. Once targeted, MLL recruits CREBBP/EP300, MOF, MOZ, HBO1, HCF and other effectors to activate transcription [24, 28, 30, 41, 42]. By analogy to trithorax, it is thought that these activities impart stability through cell division to MLL-regulated target genes. The exact biochemical activitie(s) that impart heritability of expression to downstream targets is unclear, but may be linked to bookmarking functions [44, 45], and histone methylation or acetylation activity [46, 47].

Although regulation of the clustered HOX genes may be the most conserved and well-studied feature of MLL and its homologs, there are many unique, tissue specific programs also regulated by MLL. During hematopoiesis, MLL does sustain high levels of expression of Hoxa9 and Meis1, particularly in the most primitive hematopoietic stem cells (HSCs) and multi-potent progenitors (MPPs) to promote expansion of these undifferentiated cell pools [19, 48–50]. In addition, Hlf, Eya1, Mecom, Prdm16, Pbx1 and other self-renewal regulators are likely direct targets, sustained by Mll in HSCs/MPPs [51].

Knockout of MLL specifically in adult hematopoietic stem/progenitors (HSPCs) results in a transient increase in HSC proliferation, loss of self-renewal, and decrease in hematopoietic progenitors [50, 52]. Many MLL target genes exhibit reduced expression as HSCs differentiate [19, 49, 53]. The fact that MLL levels remain fairly constant throughout differentiation suggests that an opposing chromatin modification counteracts its activity upon differentiation, or it becomes uncoupled from target genes by post-transcriptional mechanisms. This programmed silencing of Hoxa9, Meis1, Mecom and other genes is required to maintain normal hematopoiesis, as the sustained expression of these genes can lead to leukemia [54, 55].

MLL is a member of a 6-member family of histone methyltransferases that represent 3 paralog pairs (KMT2A/B, KMT2C/D, KMT2E/F). Each family member interacts with the SET domain-associated enzymatic complex described above, and each make common paralog-specific protein interactions as well as unique interactions. Whereas paralogs MLL3/4 (KMT2C/D) are more strongly associated with H3K4me1 at enhancers, MLL1/2 paralogs are more strongly associated with H3K4me2/3 at TSSs [56]. All six members are expressed ubiquitously yet perform very distinct functions depending on the cell type (reviewed in [57]). For example, the expression of MLL and MLL2 (KMT2A and KMT2B) overlap throughout the hematopoietic system yet MLL is an essential regulator of the HSC program described above, whereas MLL2 loss has no impact on the expression of these genes or on HSC function [58]. In contrast, MLL2 controls a suite of genes involved in macrophage function [59]. Similarly, whereas MLL fusion oncoproteins readily transform hematopoietic progenitors, they are rendered inactive when the MLL2 N-terminus is swapped out in place of the MLL N-terminus, despite many similar chromatin-targeting domains. Although the region around the CXXC domain was implicated, subsequent experiments using the MLL2 CXXC domain illustrated that this domain alone cannot explain the target gene specificity and biological differences between MLL/MLL2 [60, 61]. In summary, the specificity of action of each of the KMT2 family members is likely explained by combinatorial protein interactions to achieve selective gene targeting, as well as protein interactions that result in the selective transcriptional programs observed upon genetic perturbation of each of the six KMT2 family members.

Mechanisms of leukemic transformation by MLL fusions through the AEP coactivator

MLL gene rearrangements lead to many distinct fusions [12], it has been challenging to understand whether fusions share any common features. Frequencies of each fusion partner vary such that only a select few partner genes account for the majority of the MLL-rearranged leukemia cases, which include AF4 (also known as AFF1), AF9 (also known as MLLT3), ENL (also known as MLLT1), AF6, AF10 (also known as MLLT10), ELL, and the non-fusion MLL PTD (Figure 2A). The uneven frequencies of fusions in leukemia may relate to the unique protein interactors (see below), contribution to a fitness advantage such as tolerance to the reciprocal chromosomal rearrangement, or translocation opportunity through distinct 3D chromatin proximities [62].

Figure 2. Functions of the AEP complex, the major fusion partner of MLL.

A) Prevalence of fusion partners in MLL-r AML and ALL combined (reviewed by Meyer et al.) [12]. The major MLL fusion partners are the components of AF4 family/ENL family/pTEFb complex (AEP). AF4 and AF5Q31 associate with each other to form a dimer. B) The structures/functions of AF4 fused with the minimum-targeting module (MTM) composed of the PWWP domain and the CXXC domain. Associating factors for each functional module is indicated by dotted lines. Interactions/functions necessary for gene activation and leukemic transformation by the MTM-AF4 fusion are highlighted in red dotted lines. Light green spheres indicate non-constitutive associating factors that bind in a sub-stoichiometric manner. C) Assembled minimal oncogenic protein complex minimally required for leukemic transformation. D) Model of AEP-mediated gene activation. AEP recruits SL1, ELL, and MED26 non-constitutively. SL1 provides TBP to the promoter which leads to the formation of a pre-initiation complex of RNA Polymerarse II. AEP promotes transcription elongation by phosphorylating RNAPII, DSIF, and NELF to release the promoter-proximal pausing.

MLL fusions deregulate a subset of the normal MLL-dependent transcriptional program [48, 49, 63, 64], thus the basis of this altered activity is of fundamental interest. Purification of the endogenous protein complex of AF4, which is the most frequent MLL fusion partner, identified stable interaction partners [65]. AF4 associates with AF5Q31, an AF4 family protein, ENL, CDK9 and CyclinT1, forming the “AEP complex” (AF4 family/ENL family/P-TEFb complex) [65, 66] (Figure 2A, B). ENL and AF9 are homologs whereas CDK9 and CyclinT1 are known as part of the P-TEFb complex (positive transcription elongation factor b), which phosphorylates the C-terminal tail domain of RNA Polymerase II on Ser 2 to promote transcription elongation. AEP is a key target for MLL fusion mediated transformation [65] and also associates with ELL elongation factors 1/2/3, another major category of MLL fusion partners, to form a super elongation complex [66].

To identify the minimum structure required for oncogenic transformation, artificial fusion constructs composed of combinations of functional modules required for target recognition and transcriptional activation were tested for transforming ability [41, 61, 67, 68]. This approach revealed that the binding surfaces for P-TEFb or ELL elongation factors, the NHD domain and ALF domain, respectively, are not absolutely required for transformation [67] (Figure 2B). The pSER domain, which encompasses the sub-domains termed the SDE motif, DLXLS motif, and NKW motif was necessary and sufficient for transformation. This result indicated that the recruitment of elongation factors (i.e., P-TEFb and ELL) is not the driving force for transformation, but the functions mediated by the pSER domain rather plays an essential role. Interactome analysis revealed that the SDE motif is responsible for interaction with the SL1 complex composed of TAF1A/B/C/D and TBP, previously known for the transcription factor for RNA Polymerase I-mediated transcription. The NKW motif is presumably required for TBP loading for subsequent formation of the pre-initiation complex of RNA Polymerase II (Figure 2B–D). TAF1C was co-localized with the MLL fusion at target loci such as HOXA9, suggesting that the SL1 complex is critically implicated in AEP-mediated gene activation. AF4 family proteins also bind to MED26, a Mediator complex component, through its DLXLS motif [68, 69]. However, deletion of the DLXLS motif did not cause loss-of-transformation, suggesting that MED26 interaction is not critically required for oncogenic transformation [67, 68]. These observations suggest that AEP is a multi-functional factor which promotes both SL1-mediated transcription initiation and P-TEFb-mediated transcription elongation.

Direct recruitment of this multi-functional AEP transcriptional coordinator complex is the driving force of gene activation and oncogenic transformation (Figure 3A). Because MLL-AF4, MLL-ENL, and MLL-AF9 fusions bind to AEP complex components directly, they can recruit AEP constitutively. This “direct AEP recruiter type” fusions account for two-thirds of all MLL rearranged leukemia cases (Figure 2A) therefore this direct AEP recruitment mechanism may be the simplest route to transformation. ELL, which binds directly to AF4 family proteins, also acts as a direct AEP recruiter-type [41, 66].

Figure 3. Five mechanisms to achieve aberrant gene expression by classes of oncogenic MLL variants.

A) The direct AEP recruiter type represented by common translocations MLL-AF4, MLL-AF9, MLL-ENL, and the less common ELL and AF5Q31 fusions. AF9 and ENL are homologs that directly bind AF4 family proteins (e.g., AF4, AF5Q31). ELL is a non-constitutive associating factor of the AF4 family protein. B) The acetyl mark provider type, represented by AFX or p300/CBP fusions. Deposition of H3K9/18/27 acetylation results in ENL-mediated AEP recruitment. C) The ENL provider type, represented by MLL-AF10 and MLL-AF17. Direct recruitment of the DOT1L complex via the AF10 family portion brings ENL, providing it on the target chromatin to from an AEP complex. D) The multimerization type, represented by MLL-AF6 and MLL-GAS7. Concentration of associated acetyltransferases such as HBO1 result in H3K14 acetylation thus recruitment of ENL and an AEP complex. E) The partial tandem duplication type may also recruit additional acetyltransferases thus acetyl-binding AEP.

MLL undergoes fusions with AFX, CREBBP, and EP300 in rare cases. AFX is a transcriptional regulator that binds CREBBP/EP300 [70]. These MLL fusions recruit the histone acetyltransferase activity of CREBBP/EP300 to provide the acetyl marks on target promoters to aberrantly activate transcription [71] (Figure 3B). These “acetyl mark provider type” MLL fusions circumvent the normal recruitment of CREBBP/EP300 mediated by WT MLL, thus constitutively providing H3K18/27ac to target promoters [28]. The YEATS domain of ENL family proteins is a chromatin reader module that specifically recognizes H3K18/27ac marks [72, 73]. Thus, H3K18/27ac marks on chromatin recruit ENL and allow the formation of AEP on target promoters. These acetyl mark provider type MLL fusions indirectly recruit AEP to activate transcription of MLL target genes.

ENL family proteins are also a component of the DOT1L histone methyltransferase complex containing AF10 and AF17. These two proteins are also fusion partners of MLL, thus demonstrating an alternative means to recruit ENL to create fusion oncoproteins [74, 75]. MLL fusions to AF10 or AF17 recruit a DOT1L-ENL complex to MLL target loci and load ENL to target chromatin containing H3K18/27ac marks to promote AEP-mediated gene activation [76]. Interestingly, this “ENL provider type” MLL fusion requires the THD2 domain (formerly termed TRX2 domain [77]) for full-transforming capacity [76] (Figure 3C). The THD2 domain binds to the HBO1 (also known as KAT7) histone acetyltransferase [41], suggesting that HBO1-mediated acetylation on H3K14 may promote acetylation of H3K18/27 or ENL loading.

Some MLL fusions transform hematopoietic progenitors via mechanisms that remain unclear. MLL-AF6, MLL-GAS7, and MLL-GPHN do not have any of the aforementioned functions in their fusion partner portions, yet possess multimerization domains [78–80]. When these MLL fusions are transduced to murine HSPCs, typical hallmarks of MLL oncoprotein transformation are observed. In a leukemia cell line expressing MLL-AF6, AEP and di-methylated H3K79me2/3 were colocalized with MLL-AF6 at the HOXA9 promoter [65], indicating that MLL-AF6 can somehow constitutively recruit AEP and DOT1L. It is unclear how multimerization domains can confer this activity. One possibility is that this “multimerization type” MLL fusion may recruit HBO1 efficiently via paired THD2 domains, which may trigger AEP recruitment (Figure 3D).

The MLL PTD mutant represents an even more challenging case to understand. This intra-genic duplication occurs in ~5% of MLL-r leukemias [12]. This MLL rearrangement encodes two sets of CXXC and THD2 domains, which may impart stronger target gene binding than WT MLL [81]. Many proteins possess a CXXC domain that binds unmethylated CpG, implying that the MLL PTD may outcompete these proteins including non-rearranged MLL itself [82]. Through the extra THD2 domain, additional HBO1 may be recruited thus impacting downstream target gene expression [41, 83, 84] (Figure 3E).

In summary, MLL fusions can be categorized into 5 functional groups: (1) direct AEP recruiter type, (2) acetyl mark provider type, (3) ENL provider type, (4) multimerization type, and (5) partial tandem duplication type. In most of those categories, there is evidence that MLL fusions function as constitutive-active transcriptional regulators, including upregulation of HOXA9 and constitutive recruitment of the AEP complex. An intriguing possibility is that these distinct functional groups affect slightly different sets of target genes and therefore will respond differently to targeted therapies. On the other hand, it is clear that MENIN is critically required for most if not all MLL oncoproteins to bind to and stably associate with the promoter, potentially serving as an ideal therapeutic target.

Collaborative gene activation by MLL, AEP, and MOZ/MORF histone acetyltransferases

The transcriptional activation system mediated by MLL and AEP is aberrantly activated by MLL fusions. In addition to MLL and AEP, MOZ (also known as KAT6A) histone acetyl transferase and its homolog MORF (also known as KAT6B) play important roles to promote gene activation in this system. MOZ/MORF maintains HOX gene expression in development [85–89]. MOZ knockout mice cannot sufficiently produce hematopoietic progenitors partly due to reduced HOX gene expression, similar to MLL knockout mice [90, 91]. MOZ/MORF fuses with TIF2, which binds CREBBP/EP300, or CREBBP/EP300 itself to cause leukemia [92–95]. These MOZ fusions are thought to provide H3K18/27ac marks via the CREBBP/EP300-mediated HAT activity on their target promoters similar to the “acetyl provider” type MLL fusions (Figure 3B). Both MLL and MOZ bind to promoter regions containing unmethylated CpGs; MOZ and MORF possess a winged helix domain 1 (WH1) in their N-termini, which specifically recognizes unmethylated CpGs [42, 96, 97](Figure 4). Furthermore, MOZ and MORF have been shown to interact with MLL [42, 98]. MLL and MOZ significantly co-localize at CpG-rich promoters in HEK293T cells [42]. On such loci, MLL recruits CREBBP/EP300 to introduce H3K18/27ac marks and recruits HBO1 to introduce H3K14ac marks [28, 41], while MOZ binds to H3K14ac marks [99] and introduces H3K23ac marks to the same target chromatin [85]. These observations provide one mechanism for collaborative actions by MLL and MOZ/MORF.

Figure 4. Steps in the MLL–MOZ–AEP-mediated transcriptional activation system.

Perturbation by MENIN, DOT1L, and KAT6/7 inhibitors is indicated in red.

A second mechanism by which MLL collaborates with MOZ/MORF is through the ENL complex. Purification of ENL-containing complexes identified a subset that contain MOZ/MORF proteins [42]. Previous attempts to characterize the core MOZ/MORF complex did not find ENL, suggesting a sub-stoichiometric or cell context-dependent association between MOZ/MORF and ENL [42, 85]. In MOZ-knockout cells, ENL and DOT1L exhibit reduced occupancy at HOXA loci [100], suggesting that MOZ and ENL form a complex then recruit a DOT1L complex (Figure 4). Leukemic CALM-AF10 fusion utilizes the MOZ-ENL complex as a chromatin targeting factor, wherein CALM-AF10-DOT1L complex is recruited to the promoter-bound MOZ-ENL complex [100]. Recently it was reported that NUP98 fusions also interact with MOZ/MORF HATs to promote leukemic transformation [101]. These notions suggest that multiple leukemic oncoproteins utilize MOZ/MORF HATs in leukemic transformation. In summary, MOZ/MORF acetyltransferases target a highly overlapping set of target genes as compared to MLL, introduce distinct epigenetic marks, and collaborate to activate gene expression. MLL fusions and MOZ fusions both hijack a key recruitment step in this MOZ–MLL-AEP-mediated transactivation system to constitutively activate target genes to cause leukemia. This transactivation system appears to be aberrantly activated not only by MLL and MOZ fusions, but also by CALM-AF10 and NUP98 fusions, potentially serving as a common therapeutic target.

Exploiting molecular mechanisms to develop targeted therapy

Understanding the molecular mechanisms underlying leukemic transformation by MLL fusions is critical for the rational design of targeted therapies. The first approaches to disrupting ectopic activities of MLL fusion oncoproteins focused on gained interactions through the fusion partner (DOT1L catalytic function) or critical chromatin-targeting contacts such as MENIN interaction [102, 103] (Figure 4). The catalytic function of DOT1L is critically required for repelling suppressive epigenetic factors such as SIRT1 [104]. DOT1L inhibitors demonstrated some efficacy but did not progress to clinical use due to pharmacokinetic limitations. Grembecka and colleagues developed structure-guided MENIN-binding small molecules, which disrupt chromatin localization of MLL complexes and thereby prevent the recruitment of transcriptional activation complexes described above to target gene promoters [105, 106]. This results in HOXA9 and MEIS1 downregulation, leading to leukemia cell differentiation and apoptosis. Preclinical studies have demonstrated that MENIN inhibition effectively reduced leukemia burden in MLL-r cell lines and mouse models, without significantly impacting normal hematopoiesis [107, 108].

These promising preclinical findings spurred further drug development and supported the launch of multiple clinical trials (NCT04811560, NCT04065399, NCT04067336, NCT04988555), including the pivotal AUGMENT-101 and KOMET-001 trials evaluating MENIN inhibitor monotherapy for relapsed or refractory MLL-r and NPM1-mutant leukemias. NPM1-mutant leukemias were included due to the involvement of the WT MLL complex and empirical findings [109, 110]. Many additional trials have been launched based on limited toxicity and promising efficacy findings [111, 112]. Clinical trials have demonstrated that the MENIN inhibitor, revumenib, achieved a 43.9% composite complete remission (CRc) rate, meaning nearly half of treated patients showed a major reduction or disappearance of leukemic blasts. Notably, 68% of responders achieved measurable residual disease (MRD) negativity, indicating no detectable leukemia by sensitive assays — an outcome often associated with durable remission. Revumenib is now FDA-approved for MLL-r relapsed or refractory leukemia. However, a considerable proportion of patients eventually developed resistance, leading to disease relapse.

Resistance mechanisms identified from this trial, and potential mechanisms revealed in pre-clinical model systems can be broadly classified into genetic mutations and epigenetic rewiring (Figure 5), each of which would allow leukemia cells to evade MENIN inhibition while sustaining an oncogenic gene program. The resistance mechanism identified in treated patients involves acquired mutations in MEN1, the gene encoding MENIN. In the AUGMENT-101 trial, about 38% of patients who initially responded and remained on revumenib for at least two cycles acquired new somatic MEN1 mutations which were not present at baseline. The resistance-associated mutations observed clustered at MEN1 residues M327, G331, T349, and S160 [113]. The mutations specifically alter the MENIN binding pocket, reducing the binding affinity of MENIN inhibitors while preserving MENIN’s ability to interact with MLL fusion proteins and tether them to chromatin via LEDGF, thereby sustaining leukemogenic transcription (Figure 5A). By lowering inhibitor affinity by 30- to 300-fold, treatment is rendered ineffective [114]. Thus further development of new generation MENIN-MLL inhibitors that overcome resistant structural changes is needed.

Figure 5. Mechanisms of acquired resistance to MENIN-MLL inhibitors.

A. MENIN-MLL inhibitor-mediated eviction of MLL/MENIN complex from target promoters (left) and its failure by structural alteration that inhibit the drug binding while preserving MENIN-MLL-LEDGF complex formation (right). B. Epigenetic rewiring occurs following the epigenetic imbalance caused by the eviction of MLL/MENIN complexes with MENIN-MLL inhibitors. Loss-of-function mutations on the factors implicated in the rewiring lead to resistance to MENIN-MLL inhibitors. Drug resistance-induced events are indicated in red. Counteractive measures to the drug resistance are indicated in bule.

Combination therapy which can effectively eliminate leukemic clones may reduce the likelihood of drug resistance mechanisms as described above. Trials such as NCT06448013, NCT05360160, NCT07007312, and NCT06284486 add MENIN inhibitors to a backbone of venetoclax (a BCL2 inhibitor), with or without azacitidine or gemtuzumab or standard of care chemotherapy (NCT06313437, NCT05761171). One trial, (NCT06222580) co-targets MENIN and FLT3 in patients with mutant FLT3, an indirect target gene of the MENIN-MLL complex. These combinations were supported by pre-clinical data illustrating synergy between MENIN inhibitor and FLT3 or BCL2 inhibition [115–117]. Other combination trials take advantage of FDA approved drugs with good clinical track records but were not necessarily driven by specific molecular mechanisms. Additional recent pre-clinical model studies illustrate co-dependencies that may suggest more rational drug combinations.

Figure 5B illustrates additional potential mechanisms by which leukemia can evade MENIN inhibition of oncogenic gene expression. One study identified a molecular switch between the MLL1-MENIN and MLL3/4-UTX chromatin-modifying complexes. MENIN inhibitors cause an imbalance of epigenetic regulation followed by an epigenetic rewiring. Disruption of the MLL1-MENIN complex lifts its blockade on the MLL3/4-UTX complex, enabling activation of a tumor-suppressive transcriptional program essential for therapeutic response [118]. However, failure to activate MLL3/4-UTX—due to loss of its complex components— can prevent this tumor-suppressive program, contributing to resistance by allowing leukemia cells to maintain oncogenic transcription despite MENIN inhibition. Similarly, Zhou et al. identified a resistance mechanism involving non-canonical MENIN target genes, where loss of PRC1.1 components such as BCOR and PCGF1 leads to derepression of oncogenes such as MYC, overcoming MENIN inhibitor-mediated suppression of leukemia programs [119].

To cope with emerging resistance to MENIN-MLL inhibitors, a better method to effectively eliminate all leukemia cells in the first attempt is needed. To achieve that, simultaneously striking pathways directly or indirectly contributing the MLL-mediated leukemic transformation has been explored. Genome-wide screens for enhancers of MENIN inhibition illustrated that IKAROS can collaboratively regulate leukemogenic gene expression programs downstream of MLL fusion oncoproteins together with MEIS1, thus IKAROS degrader synergistically kills leukemia cells with MENIN inhibitor [120]. Similarly, MENIN inhibition collaborates with inhibition of BRD4, EP300, and HBO1 to enhance leukemia cell killing and in particular the combination with a KAT6/7 dual inhibitor, can overcome both genetic and non-genetic resistance mechanisms [115, 121](Figure 5B).

Two additional studies have demonstrated that inhibiting guanine nucleotide biosynthesis, particularly via IMPDH2 blockade, has multiple effects on MLL-r leukemia. First, inhibition of IMPDH2 and depletion of guanine nucleotides reduces rRNA transcription and selectively reduces MLL-fusion targets through cumulative impact on protein translation and oncogenic fusion complex components, rendering cells more sensitive to MENIN inhibition [122, 123]. Liu and colleagues show that IMPDH inhibition and guanine nucleotide depletion also results in activation of a TLR-NFκB pathway, leading to differentiation in vitro and reduced leukemia burden in vivo [122]. Collectively, these studies illustrate the potential for interfering with collaborative pathways to prevent the emergence of, or more effectively kill MENIN resistant MLL-r leukemia.

Despite the challenges posed by resistance, MENIN inhibitors remain one of the most promising targeted therapies for MLL-r leukemias. Ongoing clinical trials are actively investigating biomarkers that can accurately predict resistance, ensuring more personalized treatment strategies. The future of MENIN-targeted therapy likely lies in rational combination approaches that preemptively block escape pathways before resistance emerges. By integrating genomic, transcriptomic, epigenetic profiling, and adaptive treatment strategies it may be possible to achieve long-term remission in patients with aggressive leukemia subtypes such as this MLL-r group. Ultimately, the dissection of detailed molecular mechanisms driving leukemogenesis and therapeutic resistance is essential for the development of next-generation inhibitors and for ushering in an era of precision medicine for high-risk leukemias.

The development of MENIN inhibitors exemplifies how a deep mechanistic understanding can drive successful drug discovery. This effort reflects a rigorous pursuit of the molecular underpinnings of leukemia, persistent efforts to model the disease, and successful efforts to inhibit protein-protein interactions involving nuclear factors—once considered undruggable targets. This progress represents a collective triumph of both academic and industrial efforts. The story of MENIN inhibitors affirms that mechanistic insight is not only foundational to therapeutic innovation but also paves the way for future advances in the treatment of high-risk leukemias.

Highlights.

• Chromosomal translocations altering MLL (KMT2A) result in an overly active oncogenic complex including MENIN and multiple histone acetyltransferases

• MLL oncoproteins exert their oncogenic properties with the AEP coactivator via five distinct mechanisms

• The MLL–MOZ–AEP-mediated transcriptional activation mechanism is overly activated in MLL- and non-MLL-rearranged leukemia.

• Drugs that target the MENIN-MLL interaction effectively kill MLL-rearranged leukemia cells have been developed and are now in clinical use.

• Multiple mechanisms of acquired drug resistance to MENIN inhibitors pose new challenges