NUP98 and KMT2A: usually the bride rather than the bridesmaid

In human hematopoietic malignancies, KMT2A and NUP98 are each independently targeted by numerous chromosomal alterations leading to the expression of fusion oncogenes. In this issue of Haematologica, Fisher and colleagues from J. Schwaller's team report the functional study and creation of an in vivo model1 for a unique fusion between these two genes2 showing that leukemia development by NUP98-KMT2A is not associated with classical KMT2A fusion mechanisms.

KMT2A (a.k.a. MLL) is a large protein of almost 4,000 amino acids that is processed by the endopeptidase Taspase1. It interacts with numerous proteins and assembles into large protein complexes (Figure 1). The functions of KMT2A include writing the H3K4me3 chromatin mark characteristic of active promoter regions through its C-terminal SET domain. In both lymphoid and myeloid malignancies, KMT2A is targeted by numerous chromosomal alterations resulting in the expression of fusion oncogenes with over 80 different partners in toto (https://mitelmandatabase.isb-cgc.org/). Experimental models have demonstrated that several fusions containing the N-terminal portion of KMT2A and various partners [here termed KMT2A-X, where X is frequently AFF1, MLLT3, MLLT10 or MLLT1 in acute lymphoid leukemia patients, and MLLT3, MLLT10, MLLT1 or ELL in patients with acute myeloid leukemia (AML)] are important for disease development and maintenance.43

Figure 1.Molecular mechanisms associated with KMT2A-X and NUP98-KMT2A fusions. Schematic representation of the functional domains of wildtype KMT2A, KMT2A-X (where X corresponds to the fusion partners), and NUP98-KMT2A and the molecular mechanisms associated with leukemic transformation. While KMT2A-X fusions are associated with high expression (High) of HOX genes and MYC, NUP98-KMT2A is associated with low HOX genes (Low) and its regulation on MYC expression remains unknown (?). In mice, NUP98-KMT2A transformation is associated with high expression of cell cycle-associated genes (e.g. Sirt1, Rbl2 and Tert2). Therapeutic targeting developed for KMT2A-X leukemia includes MEN1, DOT1L and BRD4 inhibitors. Whether small-molecule inhibitors of SET domain or TASPASE 1 activities would be efficient in NUP98-KMT2A leukemia is unknown. AT: AT-hook domain; SNL: speckled nuclear localization domains; MBD: MENIN 1-binding domain; PHD: plant homeodomain finger domain; BD: bromodomain; FYRN and FYRC: phenylalanine/tyrosine-rich-N- and C-terminal domains; TAD: transactivation domain; SET: SET methyltransferase activity domain.

It has long been recognized that KMT2A-X fusions activate transcription of different HOX genes (e.g. HOXC8, HOXA7, HOXA9, and HOXA10) and are associated with high expression of the HOX cofactor MEIS1. At the molecular level, at least two distinct mechanisms have been involved in KMT2A-X leukemogenic properties and the deregulated expression of KMT2A-X target genes (Figure 1). On the one hand, the first 145 N-terminal amino acids of KMT2A interact with MEN1 and LEDGF to bind KMT2A target genes.5 On the other hand, most fusion partners of KMT2A belong to the transcription elongation machinery leading to the active recruitment of various factors including (i) the P-TEFb complex (comprising CDK9), which phosphorylates RNA polymerase II; and (ii) the histone methyltransferases DOT1L and NSD1, which catalyze H3K79me3 and H3K36me2 marks deposited in the body of actively transcribed genes. This creates an active gene transcription elongation environment at KMT2A-X target genes (e.g. HOX genes), which is reinforced by the recognition of acetylated lysines on histones at important oncogene loci (e.g. MYC) by the BET proteins including BRD4. Based on these dependencies, small molecule inhibitors of DOT1L, of BRD4 and of the interaction between KMT2A and MEN1 have been developed.86

Other alterations of KMT2A function are observed. In some instances reciprocal X-KMT2A fusions were shown to contribute to leukemogenesis in murine model (e.g. AFF1-KMT2A cooperation with KMT2A-AFF19). KMT2A partial tandem duplications (KMT2A-PTD) are also found in AML and both murine modeling and human genetics indicate that KMT2A-PTD requires additional mutations to induce bona fide leukemia.1110

Wildtype NUP98 is part of the nuclear pore complex, a large structure of ~30 proteins at the nuclear membrane which provides a bidirectional channel allowing small ions and peptides to diffuse and larger molecules (mRNA and proteins) to be actively transported by carriers. NUP98 is different from other nucleoporins as it contains multiple GLFG repeats allowing interaction with CREBBP/EP300 and it can be found throughout the nucleus. Nup98 was reported to be involved, together with Rae, in cell cycle progression and mitotic spindle regulation.12 Notably, NUP98 is found at sites of actively transcribed genes presenting the H3K4me3 mark and is involved in cell cycle and development.13 NUP98 is also involved with wildtype KMT2A and NSL in complexes regulating HOX gene expression.14

In leukemia, NUP98 is recurrently fused with over 30 different partners (including NSD1, KDM5A, but also homeodomain proteins such as HOXA9, HOXC11, HOXD11 or HOXD13). These fusions (termed NUP98-X here) result from chromosomal alterations that are frequently undetected by conventional cytogenetics due to the location of the NUP98 gene close to the telomere of the short arm of chromosome 11 (11p15).1615 In the case of chimeras between NUP98 and homeodomain proteins, the GLFG repeats of NUP98 generally replace the transactivation domain. To date, all NUP98-X fusions have been associated with high HOX gene expression. Similarly to wildtype NUP98, the NUP98-HOXA9 fusion also interacts with wildtype KMT2A through the second GLFG repeat of NUP98 and KMT2A is important for the recruitment of NUP98-HOXA9 to the HOXA locus and NUP98-HOXA9-dependent HOXA genes expression.17

Fisher et al. performed functional modeling of a peculiar NUP98-KMT2A alteration, resulting from an inv(11)(p15;q23) characterized in two AML patients and leading to the fusion of NUP98 exon 13 to KMT2A exon 2. The predicted NUP98-KMT2A fusion encodes a 4,340 amino acid protein lacking the MEN1-interacting domain but containing most of KMT2A including the H3K4 methylation SET domain, as opposed to KMT2A-X fusions (Figure 1). As the reciprocal KMT2A-NUP98 fusion transcript (between exon 1 of KMT2A and exon 14 of NUP98) was detected in only one of the two original patients,2 the hypothesis was that the NUP98-KMT2A fusion may represent the important disease driver. Fisher et al. achieved this tour de force through the development of a novel inducible NUP98-KMT2A transgenic mouse model.

The authors demonstrate the expansion and increased competitiveness of NUP98-KMT2A-expressing hematopoietic progenitor cells (LineageSca1Kit cells) and concomitant cell cycle abnormalities without significant changes in the relative distribution between long-term hematopoietic stem cells and multi-potent progenitors. Upon long-term NUP98-KMT2A expression, mice succumbed to lethal myelodysplasia or AML. The median latency for development of a hematopoietic malignancy in inducible NUP98-KMT2A mice was rather long (>1 year) and sublethal irradiation to generate DNA damage accelerated disease. Together with the observation of additional mutations in other human NUP98-rearranged15 or KMT2A-PTD,11 this strongly suggests that cooperating mechanisms are required for induction of full-blown AML by NUP98-KMT2A. Interestingly, however, the co-expression of Flt3-ITD did not accelerate the disease in this inducible NUP98-KMT2A model, suggesting different cooperating networks as compared to the NUP98-NSD1 fusion.18

Inducible models allow elegant and formal testing of whether continuous expression of the driver oncogene is required for leukemia maintenance by removing doxycycline treatment in diseased animals. Previously, full dependence of AML on KMT2A-MLLT3 expression was observed using a similar model.19 Here, however, ceasing the doxycycline treatment in recipients of inducible NUP98-KMT2A cells did not abrogate the disease progression in all the mice. The authors suggest that this may result from a “leak” of residual NUP98-KMT2A expression inherent to this inducible system. In this regard, given that KMT2A-X and NUP98-KMT2A fusion transcript expression is controlled by different endogenous regulatory elements - KMT2A and NUP98 promoters, respectively - significantly different levels of fusion expression could be required for leukemia induction and maintenance. It remains to be formally tested whether NUP98-KMT2A expression is essential for the maintenance of already transformed leukemic cells in all cases.

At the molecular level, inducible NUP98-KMT2A murine leukemia cells, similarly to NUP98-KMT2A patient’s leukemic cells,2 do not show significant upregulation of HOX genes as compared to control cells or leukemia from two retroviral models of KMT2A fusions (KMT2A-ENL and KMT2A-MLLT3). Gene expression analyses in transgenic mouse embryonic fibroblasts confirmed cell cycle deregulation and further demonstrated a block in induction of senescence. Notably, a subset of cell cycle- and senescence-associated genes deregulated by transgene induction in mouse embryonic fibroblasts was also found to be deregulated in murine hematopoietic progenitor cells (e.g. Sirt1, Rbl2, Tert2). These data suggest that NUP98-KMT2A does not transform hematopoietic progenitor cells through aberrant expression of HOX genes and cofactors but through an alternative mechanism associated with a defective cell cycle checkpoint. Notably, this is further supported by the absence of significant cell cycle perturbation in inducible NUP98-KMT2A cells mediated by small MEN1 or BRD inhibitors, as opposed to their effects on KMT2A-MLLT3 cells.

Three NUP98 fusion partners (NSD1, NSD3, and KMT2A) have a SET domain and another partner is a known interactor of SET-containing proteins (SETBP1) with histone methyltransferase function. Although additional genome-wide chromatin analyses will be required to assess H3K4me3 profiles and NUP98-KMT2A DNA binding sites in NUP98-KMT2A cells, it could be hypothesized that aberrant deposition of H3K4me3 at NUP98 target genes enhances or ectopically creates promoter activities. More generally, it also remains to be determined: (i) whether the NUP98 or KMT2A moiety controls the identity of the target genes; (ii) whether dimerization is required for transformation as for other KMT2A-X fusions;20 and (iii) whether the location of NUP98-KMT2A at the nuclear pore, reported to be in close proximity to the loci of actively transcribed cell cycle regulators, in part controls the identity of the target genes in a cell context-dependent manner.

Together, the identification of transcriptional targets of NUP98-KMT2A represents a first step toward the development of novel therapeutic strategies. Based on the protein structure, the NUP98-KMT2A transforming properties could depend on cleavage by TASPASE 1 and SET domain catalytic activity. As interference with these activities has been proposed,21 future assessment of the efficacy and specificity of targeted therapies could be of interest in these human leukemias.

Acknowledgments

We thank Olivier A. Bernard and Brian J. Huntly for scientific discussions.

Footnotes

• FundingAF is supported by Fondation pour la Recherche Médicale. TM is supported by Institut National Du Cancer (PLBIO-2018-169), PAIR-Pédiatrie/CONECT-AML (Collaborative Network for Children and Teenagers with Acute Myeloblastic Leukemia: INCa-ARC-LIGUE_11905 and Association Laurette Fugain), Société Française des Cancers de l'Enfant, INCa-DGOS-INSERM_12551.