Summary: Acute myeloid leukemia (AML) is now recognized to be an imprecise term that refers to a range of myeloid malignancies that have different genetical etiologies, clinical characteristics, and therapeutic sensitivities. Targeting the MLL1 and DOT1L histone modification complexes, both alone and in combination, showed activity against AML driven by a mutant NPM1 protein in several preclinical models and may represent a new treatment direction for this devastating disease. Cancer Discov; 6(10); 1087–9. ©2016 AACR.
See related article by Kühn and colleagues p. 1166.
Progress in the development of effective novel therapeutics for treatment of acute myeloid leukemia (AML) has been generally disappointing, and clinical outcomes remain generally suboptimal. The notable exception to this trend is the remarkable advances seen in acute promyelocytic leukemia (APL), with remission and long-term survival rates improving from less than 15% to greater than 90% over the past 50 years (1). These improvements occurred following the recognition of APL as a clinical entity distinct from other forms of AML, with different underlying biology and hence potential unique therapeutic sensitivities. Although mechanistic understanding of successful targeted therapy in APL occurred after clinical success, it did provide a template for how improved disease definition into subgroups based on underlying etiology rather than phenotype may provide opportunities for targeted intervention. It is now clear that APL is not a special case, and that earlier morphology and cytogenetic-based classifications did not adequately reflect the diverse genetic etiology of the broad class of myeloid malignancies labeled as AML.
Technological advances over the past decade have allowed high-throughput DNA sequencing of thousands of primary AML patient samples (2). This has revealed a catalog of recurrent somatic mutations in fewer than 80 genes or genomic regions, which can be used to segregate patients with AML into one of approximately one dozen subgroups with potentially differing predictive and prognostic significance. In the recently updated World Health Organization diagnostic classification, AML with mutated nucleophosmin (NPM1mut) is now recognized as a distinct subset, with the presence of this somatic mutation now superseding traditional morphologic descriptors such as multilineage dysplasia (3). NPM1 mutations are found in 25% to 30% of patients with AML, making it the second most common recurrent somatic mutation and the distinguishing feature of the largest genomic subgroup (2).
NPM1 ordinarily functions as an intracellular chaperone, and when mutated is characterized by aberrant cytoplasmic localization detectable by immunohistochemistry. Although greater than 50 variants of NPM1 mutation have been described, almost all are four-base pair insertions within exon 12, which lead to a frameshift and premature truncation of the NPM1 protein. NPM1 mutations are found in adult patients with AML of all ages, but are less common in children, particularly those under 3 years of age (4). Patients with NPM1mut AML commonly have co-mutated DNMT3A and/or FLT3-ITD and are associated with a characteristic gene expression profile including downregulation of CD34 and HOX gene upregulation. Mutations in NPM1 are typically seen in association with normal karyotype AML (NK-AML); approximately 85% of NPM1mut AML is NK-AML and more than 50% of NK-AML has an NPM1 mutation (5).
Despite the clear clinical relevance of NPM1mut as a potential target for AML therapy, its mechanism of action remains incompletely understood. Several lines of investigation have suggested that enforced expression of HOX genes, especially HOXA7, HOXA9, and HOXA10 (hereafter referred to as the “HOXA cluster”), and the HOX binding factor MEIS1 are important for the oncogenic ability of NPM1mut protein, and it has been speculated that cytoplasmic sequestration of proteins by NPM1mut indirectly leads to enforced expression of the HOXA cluster and MEIS1 (6). The HOXA cluster has been shown to be an important regulator of myeloid differentiation in mammals; these genes are highly expressed in hematopoietic stem and progenitor cells, and their expression becomes extinguished as cells differentiate to monocytes and neutrophils. Of note, several other recurrent chromosomal abnormalities associated with AML in humans, including MLL1 translocations, NUP98 translocations, and monosomy 7, are associated with overexpression of the HOXA cluster and MEIS1, suggesting that enforced expression of these genes may be a common pathway associated with AML. Indeed, gene expression profile studies have shown that this HOXA cluster is overexpressed in approximately half of all patients with AML, and HOXA cluster overexpression is mutually exclusive with favorable cytogenetic [(inv 16), t(8;21), or t(15;17)] features (6).
Kühn and colleagues (7) predicted that wild-type (WT) MLL1 might be a critical regulator of HOXA cluster expression in the context of NPM1mut AML. They used CRISPR/Cas9 reagents to perform a structure/function analysis of MLL1 in the context of NPM1mut AML and found that the menin–MLL1 binding surface was required for the leukemogenic effect of NPM1mut. Realizing that this finding suggested an indirect mechanism to target NPM1mut, the authors proceeded to show that a small-molecule inhibitor of MLL1–menin binding (designated MI-503) could downregulate expression of HOXA cluster genes and MEIS1. This gene downregulation was accompanied by growth inhibition and increased differentiation of the AML cells. MEIS1 has previously been shown to be an important upstream regulator of FLT3 (8), and the authors showed that FLT3 expression was also suppressed by treatment with MI-503 (7). Using both xenotransplant and NPM1mut knock-in models, the authors showed that MI-503 treatment was also effective in vivo; this treatment was also associated with loss of menin binding at the HOX and MEIS1 loci (7).
DOT1L is an H3K79 histone methyltransferase that is known to interact with MLL1 fusion partners such as AF9 and AF10; this interaction is important for oncogenic misregulation of HOXA cluster gene expression in cells with MLL1-fusions (9). Moreover, DOT1L histone methyltransferase activity is critical for proper regulation of HOXA cluster gene expression in WT hematopoietic cells. The authors had previously shown that DOT1L inhibitors could decrease the oncogenic potential of MLL fusion proteins by decreasing HOXA and MEIS gene expression. Because the enforced expression of HOXA and MEIS1 in NPM1mut leukemic cells was associated with H3K79 di/trimethylation, the authors tested the potential for a small-molecule inhibitor of DOT1L, designated EPZ4777, to decrease leukemic cell growth and HOX/MEIS1 gene expression in NPM1mut AML cells (7). Treatment of an NPM1mut knock-in model in vivo showed a survival advantage accompanied by myeloid differentiation in the treated mice, similar to the findings with the menin–MLL1 inhibitor.
Combination of the menin–MLL1 inhibitor (MI-503) and DOT1L inhibitor (EPZ4777) led to synergistic suppression of HOXA/MEIS1 expression and growth, accompanied by monocytic and myeloid differentiation of NPM1mut cells. In vivo, the combination of MI-503 and EPZ4777 led to increased survival and decreased numbers of leukemia-initiating cells. The downregulation of MEIS1 seen in these experiments was associated with a marked decrease in FLT3 expression; this finding is consistent with the observation that FLT3 has been shown to be a direct transcriptional target of MEIS1 (8). The concomitant targeting of menin–MLL1 binding and DOT1L may enable the simultaneous indirect targeting of both NPM1mut and FLT3 (Fig. 1); this approach is of potential clinical relevance given that combined mutations of NPM1 and FLT3 are common in patients with AML and predict an unfavorable outcome. The authors also showed that several HOXB cluster genes important for normal or malignant hematopoiesis, such as HOXB2, HOXB3, HOXB4, and HOXB8, were downregulated by one or both of these agents.
These experiments highlight the potential of targeting HOX and MEIS1 expression by indirect means. This strategy is important because HOX and MEIS1 proteins have been regarded as difficult targets, given their hydrophobic binding surfaces. Moreover, the HOXA cluster genes, and to a lesser extent, the HOXB cluster, are often coordinately regulated in patients with AML (6), suggesting that targeting a single HOX protein may not be effective due to a functional redundancy among these HOX genes. For instance, although HOXA9 is the HOX gene most commonly overexpressed in leukemia, an MLL–AF9 fusion can transform Hoxa9−/− hematopoietic cells, indicating that HOXA9 expression is not required for leukemic transformation (10). The experiments by Kuhn and colleagues suggest a clever means to circumvent this problem by attacking a critical regulator upstream of HOX and MEIS1 genes.
A number of important questions remain unanswered. NPM1mut expression is associated with enforced expression of HOXA/B genes and MEIS1; the current study demonstrates that menin–MLL1 interaction as well as DOT1L histone methyltransferase activity is important for this phenomenon. However, the mechanism by which NPM1mut expression leads to HOXA/B and MEIS1 expression remains unknown. Similarly, the mechanism by which DOT1L inhibition leads to decreased HOXA/B and MEIS1 expression in NPM1mut leukemia cells is unclear. In contrast to MLL–AF9 or MLL–AF10 fusions, in which a direct interaction between DOT1L and AF9 or AF10 has been demonstrated, no direct interaction between DOT1L and WT MLL1 has been demonstrated. Moreover, the in vivo experiments performed by Kühn and colleagues to demonstrate synergism between the two agents used ex vivo pretreatment of NPM1mut cells, presumably due to unfavorable pharmacokinetic or toxicity profiles. Ultimately, this hurdle will need to be cleared before early-phase clinical trials using this combination can begin. Finally, NPM1mut AML may be associated with a variety of co-mutated genes, and the combinational impact of additional mutations on targeted therapy remains to be elucidated. Despite these questions and caveats, this indirect approach for targeting a very common clinical abnormality remains an attractive goal for treatment of patients with NPM1mut-positive AML.
Disclosure of Potential Conflicts of Interest
C.S. Hourigan reports receiving a commercial research grant from Sellas Life Sciences Group AG. No potential conflicts of interest were disclosed by the other author.
This work was supported by the Intramural Research Program of the NCI (Grants ZIA SC 010378 and ZIA BC 010982 to P.D. Aplan) and the National Heart, Lung and Blood Institute of the NIH (Grant ZIA HL 006163 to C.S. Hourigan).