Role of the RUNX1‐EVI1 fusion gene in leukemogenesis

Abstract

RUNX1‐EVI1 is a chimeric gene generated by t(3;21)(q26;q22) observed in patients with aggressive transformation of myelodysplastic syndrome or chronic myelogenous leukemia. RUNX1‐EVI1 has oncogenic potentials through dominant‐negative effect over wild‐type RUNX1, inhibition of Jun kinase (JNK) pathway, stimulation of cell growth via AP‐1, suppression of TGF‐β‐mediated growth inhibition and repression of C/EBPα. Runx1‐EVI1 heterozygous knock‐in mice die in uteri due to central nervous system (CNS) hemorrhage and severe defects in definitive hematopoiesis as Runx1–/– mice do, indicating that RUNX1‐EVI1 dominantly suppresses functions of wild‐type RUNX1 in vivo. Acute myelogenous leukemia is induced in mice transplanted with bone marrow cells expressing RUNX1‐EVI1, and a Runx1‐EVI1 knock‐in chimera mouse developed acute megakaryoblastic leukemia. These results suggest that RUNX1‐EVI1 plays indispensable roles in leukemogenesis of t(3;21)‐positive leukemia. Major leukemogenic effect of RUNX1‐EVI1 is mainly through histone deacetyltransferase recruitment via C‐terminal binding protein. Histone deacetyltransferase could be a target in molecular therapy of RUNX1‐EVI1‐expressing leukemia. (Cancer Sci 2008; 99: 1878–1883)

The t(3;21)(q26;q22) chromosomal translocation occurs in patients with aggressive transformation of myelodysplastic syndrome (MDS) or chronic myelogenous leukemia (CML). The presence of this chromosomal translocation indicates poor prognosis.^{( 1} , ² , ^{3 )} In the chromosomal joining region of t(3;21)(q26;q22), the RUNX1 gene on 21q22 is fused with the EVI1 (ecotropic viral integration site‐1) gene on 3q26.5. Previously, we cloned the RUNX1‐EVI1 fusion gene from a case with blastic crisis of CML developing additional to the t(3;21) translocation.^{( 4 )} This chimeric transcription factor is believed to be a molecular culprit for the leukemic progression of stem cell malignancies caused by t(3;21)(q26;q22).

Molecular and biological function of wild‐type RUNX1

The RUNX1 protein mainly consists of two functional domains; the Runt homology domain (which is known as a DNA‐binding domain), and the proline‐, serine‐ and threonine‐rich (PST) region (which is known as a putative transcriptional activation domain) (Fig. 1). RUNX1 forms a heterodimeric active transcriptional complex with the non‐DNA‐binding β subunit (CBFβ/PEBP2β) and binds to a specific DNA consensus sequence (ACCRCA) named PEBP2.^{( 5} , ⁶ , ⁷ , ⁸ , ^{9 )} Runx1‐ or Cbfβ‐deficient mice are embryonic lethal at day 12.5 of gestation (E12.5), showing massive hemorrhage in the central nervous system (CNS) and lack of hematopoiesis in the fetal liver.^{( 10} , ¹¹ , ¹² , ¹³ , ^{14 )} A recent study demonstrated that inactivation of Runx1 in adult mice results in megakaryocyte maturation arrest, block in T‐ and B‐lymphocyte development and increase in hematopoietic precursor cells.^{( 15 )} A subsequent study by the same group showed that the number of quiescent hematopoietic stem cells (HSC) is also negatively regulated by RUNX1.^{( 16 )} Clinically, mutations in RUNX1 have been identified in 15% to 40% of MDS‐refractory anemia with excess of blasts (RAEB) and MDS/acute myelogenous leukemia (AML).^{( 17} , ^{18 )} Patients with MDS/AML with the RUNX1 mutations have a significantly worse prognosis than those without them.

Figure 1.

Schematic structure of wild‐type RUNX1, EVI1 and RUNX1‐EVI1 molecules. Wild‐type RUNX1 has Runt homology domain at the N‐terminus and proline‐, serine‐ and threonine‐rich (PST) region at the C‐terminus. In RUNX1‐EVI1, RUNX1 protein is truncated at the end of the Runt homology domain and followed by almost the entire coding region of EVI1 molecule including the PRDI‐BF1‐RIZ1 (PR) domain.

Molecular and biological function of wild‐type EVI1

The Evi1 gene was initially identified as a frequent retrovirus integration site in myeloid tumors in AKXD mice.^{( 19 )} EVI1 expression is low in normal hematopoietic cells, but it is highly expressed in some patients of MDS or AML.^{( 20} , ^{21 )} EVI1 has four major functional domains; the two zinc finger (ZF) domains, the C‐terminal binding protein (CtBP)‐binding site, and the acidic domain (AD) (Fig. 1). The CtBP‐binding site is located between the two ZF domains, and the AD is located in the most C‐terminus. EVI1 is reported to interfere with transforming growth factor‐β (TGF‐β) signaling, and antagonize its growth inhibitory effect through targeting an intracellular signal transducer Smad3.^{( 22 )} EVI1 is also known to enhance AP‐1 activity^{( 23 )} and to block c‐Jun N‐terminal kinase (JNK) activity,^{( 24 )} as described below. Evi1–/– mice die at E10.5, and HSC in Evi1–/– embryos are markedly decreased in numbers, with defective self‐renewing proliferation and repopulating capacity.^{( 25 )} The study also shows that Evi1 directly regulates the transcription of Gata‐2, which controls both the maintenance and proliferation of HSC. It was also recently reported that the decreased colony forming capacity of Evi1–/– para‐aortic splanchnopleural (P‐Sp) cells could be rescued by retroviral Gata‐2 expression, and by blocking of TGFβ signaling using an in vitro cultivation system of murine P‐Sp regions.^{( 26 )}

The MDS1 gene was originally identified as a distinct gene from the EVI1 gene, and the putative intergenic splicing form between MDS1 and EVI1 found in normal and leukemic cells was designated as MDS1/EVI1.^{( 27 )} Because the breakpoint in the chromosome 3 of t(3;21)(q26;q22) is located in the 5′ region of the MDS1 gene, and the resultant RUNX1‐EVI1 fusion gene includes the first and second exons of the MDS1 gene between RUNX1 and EVI1 sequences, RUNX1‐EVI1 is also called as RUNX1‐MDS1/EVI1. However, at present, MDS1 is recognized as one of the alternative spliced form of the EVI1 gene.^{( 28 )} In MDS1/EVI1, a 188 amino‐acid region called the PR domain, which consists of the first and second exons of MDS1 and the second and third exons of EVI1, is fused to the N‐terminal of the EVI1 molecule. The PR domain was originally identified in human retinoblastoma protein‐binding protein RIZ1^{( 29 )} and the human transcriptional repressor PRDI‐BF1,^{( 30 )} and has been found in at least 17 kinds of diverse proteins to date. A large body of evidence suggests that PR domain family (PRDM) proteins are involved in regulation of cellular growth as well as tumorigenesis.^{( 28 )} It is interesting that the PR domains are generally located in the N‐terminal region of proteins, and that alternative splicing creates two forms of the PR domain proteins; a long form that has the PR domain (PR+) and a short form that lacks the PR domain (PR–). Decreased expression of PR+ molecules and/or overexpression of PR– molecules are observed in a variety of cancer cells, suggesting a functional antagonism, in which the PR+ forms contribute to tumor suppression while the PR– forms are oncogenic.

For the EVI1 gene, EVI1 (PR‐) is highly expressed in cases with human AML or MDS as a consequence of chromosomal rearrangements involving 3q26. Increased expression of EVI1 (PR‐) is also observed without 3q26 abnormalities. Recent studies have shown that increased expression of EVI1 (PR‐) in AML, which occurs in approximately 10% of the cases, is associated with unfavorable outcomes.^{( 20} , ^{21 )}

Molecular structure and function of RUNX1‐EVI1

Molecular structure of RUNX1‐EVI1.  The RUNX1‐EVI1 fusion gene is translated in frame to generate a chimeric transcription factor in which RUNX1‐EVI1 chimeric complementary DNA (cDNA), an open reading frame of 4185 nucleotides, encodes a 1395 amino‐acid protein. The N‐terminal of RUNX1 molecule including its Runt DNA‐binding domain is fused to almost the entire portion of EVI1 molecule (Fig. 1). Therefore, the RUNX1‐EVI1 fusion protein is a chimeric transcription molecule that consists of the Runt domain of RUNX1 and two ZF domains, CtBP‐binding site and AD of EVI1.

Dominant‐negative function over wild‐type RUNX1.  RUNX1‐EVI1 dominantly suppresses the transactivation capacity of RUNX1 through the PEBP2 sites (Fig. 2).^{( 5} , ⁶ , ⁷ , ⁸ , ^{9 )} Competition for the PEBP2 site‐binding is proposed to be a mechanism of such dominant negative effects, since RUNX1‐EVI1 binds to the PEBP2 sites with higher affinity than RUNX1 does. On binding to the PEBP2 site, RUNX1‐EVI1 is believed to recruit corepressor complex via CtBP, since it is reported that RUNX1‐EVI1 requires interaction with CtBP to repress RUNX1‐induced transactivation.^{( 31 )} In addition, the association with CtBP is also required for RUNX1‐EVI1 to block myeloid differentiation of 32Dcl3 cells induced by granulocyte colony‐stimulating factor (G‐CSF), indicating that the association with CtBP is critical for RUNX1‐EVI1 to exert its biological functions in vivo. Taken together, it is suggested that one of the mechanisms for RUNX1‐EVI1‐meidated leukemogenesis is the dominant‐negative effects over wild‐type RUNX1 through an aberrant recruitment of a transcriptional corepressor complex.

Figure 2.

(a) RUNX1‐EVI1 exerts dominant‐negative effects over wild‐type RUNX1. RUNX1 forms an active transcription factor by heterodimerizing with CBFβ through Runt homology domain. CBFβ increases the DNA‐binding ability of RUNX1 and protects RUNX1 from degradation. (b) Because RUNX1‐EVI1 binds CBFβ more strongly than RUNX1, RUNX1‐EVI1 competes RUNX1 out from the PEBP2 sites. RUNX1‐EVI1 associates with CtBP and recruits HDAC. RUNX1‐EVI1 actively represses the transcription of the RUNX1 target genes.

Stimulation of cell growth through AP‐1.  The transcription factor AP‐1 (Fos/Jun heterodimer or Jun/Jun homodimer) represents a prototype of regulatory protein that converts extracellular signals into gene expressions. AP‐1 is activated by growth stimuli, including growth factors, phorbol esters such as 12‐0‐tetradec‐anoylphorbol 13‐acetate (TPA), and various transforming oncogene products. In addition, AP‐1 functions as a positive or negative regulator in a variety of cellular differentiation processes, such as changes in the differentiation processes of embryonal carcinoma cell lines, preadipocytes and myoblasts. Previously, we showed that Rat1 cells expressing RUNX1‐EVI1 chimeric protein formed macroscopic colonies in soft agar,^{( 32 )} indicating that RUNX1‐EVI1 has oncogenic potential. Removal of the second ZF domain within the EVI1 sequence totally abrogated the ability of RUNX1‐EVI1 to transform Rat1 cells. We showed that the transforming effect is correlated with the AP‐1 activation induced by RUNX1‐EVI1.^{( 33 )}

Inhibition of JNK pathway.  JNK is activated preferentially by extracellular stress stimuli including UV light, γ‐radiation, osmotic shock, protein synthesis inhibitors, tumor necrosis factor‐α and interleukin‐1.^{( 34 )} The JNK pathway is thought to play an important role in triggering apoptosis in response to cellular stresses. The activated JNKs translocate into the nucleus where they phosphorylate transcription factors such as c‐Jun, and strongly augment their transcriptional activity. We showed that EVI1 physically interacts with and thereby inhibits the function of JNK.^{( 24 )} For instance, EVI1 inhibits stress‐induced cell death by inhibiting JNK. This inhibition of cell death by Evi1 could contribute to oncogenic transformation of cells. It has not been determined whether RUNX1‐EVI1 possesses the similar antiapoptotic effect as EVI1 does. However, given that this antiapoptotic function is dependent on the first ZF domain in EVI1, it is reasonable to speculate that RUNX1‐EVI1 also inhibits cell apoptosis by inhibiting JNK.

Suppression of TGF‐β ‐mediated growth inhibition.  TGF‐β is one of the best characterized members of growth inhibitory factors. TGF‐β inhibits proliferation of a wide range of cell types including epithelial, endothelial and hematopoietic cells.^{( 35} , ^{36 )} Intracellular mechanisms that transmit TGF‐β signaling have been clarified in detail. When TGF‐β binds to its receptors, the down‐stream signaling molecules Smad2 and Smad3 are phosphorylated by the activated TGF‐β receptors, followed by oligomerization with Smad4. The Smad2/4 or Smad3/4 complexes accumulate in the nucleus, interact with DNA, and activate the transcription of TGF‐β‐responsive genes.^{( 37} , ^{38 )} This process is apparently simple, but many proteins, including inhibitory Smads, participate in regulating the process and modify cellular responses to the stimuli. Previously we reported that EVI1 antagonizes the growth‐inhibitory effects of TGF‐β in the epithelial cells.^{( 22 )} Consistent with this finding, RUNX1‐EVI1 also represses TGF‐β‐mediated growth inhibition of a murine hematopoietic precursor cell line, 32Dcl3.^{( 39 )} The ability of RUNX1‐EVI1 to repress TGF‐β signaling depends on the two separate regions of the EVI1 portion, one of which is the first ZF domain. RUNX1‐EVI1 interacts with Smad3 through this domain, and represses the Smad3 activity as EVI1 does.

Recently we have demonstrated that EVI1 represses Smad‐induced transcription by recruiting CtBP as a corepressor,^{( 40 )} indicating that RUNX1‐EVI1 represses Smad3 activity through recruitment of CtBP.^{( 31 )} EVI1 and RUNX1‐EVI1 associate with CtBP through one of the consensus binding motifs, and this association is required for efficient inhibition of TGF‐β signaling. A specific inhibitor for histone deacetylase (HDAC) alleviates EVI1‐mediated repression of TGF‐β signaling, suggesting that HDAC is involved in transcriptional repression by EVI1. The association with CtBP is required for RUNX1‐EVI1 to block myeloid differentiation of 32Dcl3 cells induced by G‐CSF.

Several chimeric proteins, such as RUNX1‐ETO,^{( 41} , ^{42 )} PML‐RARα^{( 43 )} and BCR‐ABL^{( 44 )} have self‐interaction domains that are critical for their oncogenic capacity. RUNX1‐EVI1 was reported to homo‐oligomerize through at least three oligomerization regions,^{( 45 )} i.e. the Runt domain, the first and the second ZF domains. Deletion of the second ZF domain significantly reduces the differentiation block of primary murine bone marrow progenitors by RUNX1‐EVI1. A point mutation that inhibits CtBP binding also completely abrogates the effects of RUNX1‐EVI1 on differentiation. These results imply the importance of homo‐oligomerization for RUNX1‐EVI1 chimeric proteins to exert its leukemogenic potentials.

Repression of C/EBPα.  C/EBPα (CCAAT/enhancer binding protein α) is a leucine zipper transcription factor that regulates the expression of target genes containing C/EBP sites in their promoter regions.^{( 46} , ^{47 )} Of the genes related to hematopoiesis, these genes include CEBPA itself,^{( 48 )} CEBPE ^{( 48} , ^{49 )} and granulocyte colony‐stimulating factor (G‐CSF) receptor. ^{( 48} , ^{50 )} Inducible expression of C/EBPα is sufficient to trigger terminal granulocytic differentiation^{( 51} , ⁵² , ⁵³ , ^{54 )} and blocks monocytic differentiation program.^{( 51} , ^{53 )} Conversely, inactivation of C/EBPα in Cebpa knock‐out mice showed profound defects in the granulocytic differentiation, while all other hematopoietic cells are present in normal number, indicating its critical role in the granulopoiesis.^{( 55 )} Several lines of evidence suggest that mutations in CEBPA is one of the major molecular events that could cause myeloid malignancies. Ten per cent of patients with AML M1 or M2, according to the French‐American‐British (FAB) classification, without frequent cytogenetic abnormalities such as t(8;21)(q22;q22), carry heterozygous CEBPA gene mutations that result in truncated proteins with a dominant negative effects.^{( 49} , ⁵⁶ , ^{57 )} The RUNX1‐ETO fusion protein generated by the t(8;21) translocation represses the transcription of CEBPA by suppressing its auto‐regulatory loop in gene transcription,^{( 48 )} while the PML‐RARα fusion protein generated by t(15;17)(q21;q22) inhibits the function of C/EBPα.^{( 58 )} We confirmed that RUNX1‐EVI1 suppressed the differentiation of LG‐3 cells that differentiate along the myeloid lineage upon exposure to G‐CSF.^{( 59 )} Coexpression of C/EBPα restored the differentiation ability of the RUNX1‐EVI1‐expressing LG‐3 cells. We also found that RUNX1‐EVI1 associates with C/EBPα. RUNX1‐EVI1 suppresses C/EBPα‐mediated transcription of the CEBPA promoter in a CtBP‐binding site‐dependent fashion. In a gel‐shift assay, RUNX1‐EVI1 down‐regulated the DNA‐binding activity of C/EBPα. Therefore, the recruitment of HDAC by RUNX1‐EVI1 and interference with the DNA binding of C/EBPα could be the mechanisms for the repression of C/EBPα by RUNX1‐EVI1. These results indicate that inhibition of C/EBPα is related to the leukemogenic potential of RUNX1‐EVI1. Helbling et al.^{( 60 )} have reported that RUNX1‐EVI1 represses C/EBPα in a different way. They reported that RUNX1‐EVI1 reduces the level of C/EBPα protein but not the level of its mRNA in U937 cells and in leukemic blasts of patients carrying the RUNX1‐EVI1 translocation, and that RUNX1‐EVI1 up‐regulates the expression of calreticulin, a putative inhibitor of CEBPα translation. Calreticulin has calcium storage and chaperone function, and is postulated to be involved in the development of leukemia.^{( 61 )} The small interference RNA (siRNA) against calreticulin showed that RUNX1‐EVI1 inhibited C/EBPα expression in a post‐transcriptional mechanism through calreticulin.

Biological function of RUNX1‐EVI1

RUNX1‐EVI1 transforms the most immature hematopoietic cells.  As described above, Senyuk et al. showed that RUNX1‐EVI1 transforms primary murine bone marrow progenitors, depending on both of the CtBP‐binding site and the second ZF domain.^{( 45} , ^{62 )} Recently, Takeshita et al.^{( 63 )} introduced RUNX1‐EVI1 and its mutants in murine bone marrow cells and evaluated their transforming activities by colony replating assays. The transforming activity of RUNX1‐EVI1 was lost when any of the known functional domains of EVI1, the first and the second ZF domains, AD at the C‐terminus or CtBP‐binding site, was deleted from the chimeric protein. Although RUNX1‐EVI1 is known to repress function of wild‐type RUNX1 dominantly, forced expression of EVI1 did not transform the Runx1–/– bone marrow cells, indicating that the existence of RUNX1‐EVI1 means more than a simple combination of the presence of EVI1 and the absence of RUNX1. Interestingly, unlike the MLL‐ENL or RUNX1‐ETO leukemia‐related chimeric proteins, which transform hematopoietic progenitors, RUNX1‐EVI1 transforms only the hematopoietic stem cell fraction (c‐kit + Sca‐1 + Lin–). Moreover, RUNX1‐EVI1‐transformed cells show a cell‐marker profile distinct from that of the cells transformed by RUNX1‐ETO, which also suppresses RUNX1 function. The nature of RUNX1‐EVI1‐leukemia as hematopoietic stem cell tumors might be a consequence of these oncogenic preference of RUNX1‐EVI1.

A bone marrow transplantation model.  Cuenco et al.^{( 64 )} analyzed the effect of the human RUNX1‐EVI1 fusion gene in mouse bone marrow cells using retroviral transduction system. Mice transplanted with RUNX1‐EVI1‐expressed bone marrow cells developed acute leukemia 5–13 months after the transplantation. The disease can be transferred into secondary recipient mice with a much shorter latency period. Morphological analysis of peripheral blood and bone marrow smears demonstrated the presence of myeloid blast cells and immature cells differentiated into both myelocytic and monocytic lineages. Cytochemical and flow cytometric analysis confirmed that these mice had a disease recapitulating human AML. These observations indicate that the expression of RUNX1‐EVI1 can induce AML in mice, with extended latency period suggesting a requirement for additional perturbations. A study by the same group has demonstrated a cooperation of BCR‐ABL and RUNX1‐EVI1 in blocking myeloid differentiation and rapid induction of AML in mouse model.^{( 65 )} The study showed that RUNX1‐EVI1 alone does not block myeloid differentiation in the mouse bone marrow during the 4 months of preleukemia stage, while coexpression of BCR‐ABL and RUNX1‐EVI1 can block myeloid differentiation and induce AML rapidly. They also showed that both RUNX1 and EVI1 portions are required for RUNX1‐EVI1 to cooperate with BCR‐ABL in the induction of AML in mice.^{( 66 )}

Recently, it has been shown in a virus transduced experiment that mice transplanted with bone marrow cells expressing RUNX1 mutants developed MDS/AML within 4–13 months.^{( 67 )} Interestingly, the analysis of the viral integration sites showed that EVI1 seemed to be a collaborating gene for the RUNX1(D171N) mutant for the induction of MDS/AML. The disease has common phenotype characterized by marked hepatosplenomegaly, myeloid dysplasia, leukocytosis and biphenotypic surface markers. The collaboration between RUNX1(D171N) and EVI1 was confirmed by a bone marrow transplant (BMT) model, where coexpression of RUNX1(D171N) and EVI1 induced acute leukemia of the same phenotype with much shorter latency. These results suggest that a combination of dominant‐negative effect over RUNX1 and the oncogenic property of EVI1, both of which are components of RUNX1‐EVI1, could cause MDS or MDS/AML without additional hits. It seems important that RUNX1 should not be inactivated completely in order to cooperate with EVI1, because expression of EVI1 did not transform the Runx1‐deleted murine bone marrow cells.^{( 63 )}

Runx1‐EVI1 knock‐in mice.  We knocked‐in the Runx1‐EVI1 chimeric gene into the mouse Runx1 genomic locus to explore the effect of Runx1‐EVI1 in developmental hematopoiesis in vivo.^{( 68 )} Our knock‐in expression of Runx1‐EVI1 fusion gene results in embryonic lethality between E12.5 and E14.5, with CNS hemorrhage and a lack of fetal liver hematopoiesis. Post‐enucleated erythrocytes were absent in the peripheral blood from E12.5 Runx1‐EVI1/+ embryos, whereas nucleated erythroblasts were abundantly observed. These findings indicate that Runx1‐EVI1/+heterozygous mice fail to establish definitive hematopoiesis in the fetal liver as Runx1–/– mice. Therefore, RUNX1‐EVI1 was first demonstrated to have in vivo dominant inhibitory effects over RUNX1. Electron microscopic examination of the E13.5 fetal liver showed that a fewer number of erythroid and myeloid progenitors, and dysplastic megakaryocytes that were defective for demarcation membrane exist in the Runx1‐EVI1/+ fetal liver.

On in vitro hematopoietic colony forming assays, the fetal liver from E13.5 Runx1‐EVI1/+embryos contained multilineage hematopoietic progenitors, while that from E12.5 Runx1–/– embryos was reported to contain no definitive hematopoietic progenitors.^{( 10} , ^{11 )} No erythroid colonies were seen in the livers at both E12.5 and E13.5. The CFU‐GEMM‐derived colonies from E13.5 Runx1‐EVI1/+embryos included few erythroblasts, numerous agranular granulocytes with a delayed differentiation, and dysplastic megakaryocytes. Moreover, serial in vitro replating assays showed higher self‐renewal capacity of hematopoietic progenitors in E13.5 Runx1‐EVI1/+ fetal livers than that in wild‐type fetal livers.

On semiquantitative reverse transcriptase‐polymerase chain reaction (RT‐PCR) method with fetal livers, the expression of Pu.1 mRNA in Runx1‐EVI1/+fetal liver cells was comparable with that in the wild‐type cells, whereas its expression level was markedly decreased in Runx1–/– fetal liver cells, as has been reported previously.^{( 69 )} The maintained expression of PU.1 in Runx1‐EVI1/+fetal liver cells may support the survival of their multilineage hematopoietic progenitors up to E13.5, and enhance their monocyte/macrophage lineage differentiation. These differences in gene expression pattern provide logical explanation for the distinct hematopoietic capacity between Runx1–/– and Runx1‐EVI1/+ fetal liver cells. These data suggest that the sufficient expression of the Pu.1 gene is the prerequisite for definitive hematopoiesis in the fetal liver.

Runx1‐EVI1 knock‐in chimeric mice.  Of the six chimeric mice created, one mouse which died at 5 months of age showed marked hepatosplenomegaly. Wright‐Giemsa staining of stump preparation from the enlarged spleen demonstrated massive infiltration of large dysplastic cells, some of which contained multilobulated nuclei with various size of cytoplasm reminiscent of megakaryoblastic leukemia.^{( 70 )} Histology section showed disrupted gross architecture of the spleen, with white and red pulp intermingling, and the electron microscopic analysis of the infiltrating cells in the spleen showed 20% of the cells positive for platelet‐peroxidase, indicating that this chimeric mouse developed megakaryoblastic leukemia. The important aspect of our observation is that RUNX1‐EVI1 protein could be leukemogenic per se, in contract to RUNX1‐ETO, which requires additional hits to induce leukemia.^{( 70} , ⁷¹ , ^{72 )} This clear difference in the pathophysiological outcome likely arises from the EVI1 portion of RUNX1‐EVI1 protein. Oncogenic functions of EVI1 described above may cause stronger oncogenic capacity of RUNX1‐EVI1 than RUNX1‐ETO. Another important aspect is that the affected lineage in human leukemia is recapitulated in the experimental animal. Our observation indicates strong causal relationship between the expression of RUNX1‐EVI1 protein and megakaryoblastic leukemia. Indeed, the RUNX1‐EVI1 chimeric gene was isolated from a patient developing megakaryoblastic crisis in chronic myelocytic leukemia with additional chromosome t(3;21).

Clinical aspects

As mentioned above, the major leukemogenic effect of RUNX1‐EVI1 is through HDAC‐recruitment via CtBP (Fig. 3). Thus repression of HDAC activity is thought to suppress oncogenic effect of RUNX1‐EVI1. HDAC inhibitors (HDACi) are the member of a new class of chemical agents that epigenetically modulate gene transcription by enhancing the acetylation of core nucleosomal histones and are expected to be promising anticancer agents against various types of tumors. Suberoylanilide hydroxamic acid (vorinostat), one of the newly synthesized HDACi, is the most clinically successful HDACi that controls cutaneous T‐cell lymphoma (CTCL) effectively.^{( 73 )} Conventional HDACi, such as valproic acid and butylate, are also reported to be effective in treating some types of hematological malignancies in clinical trials. Recently, we have reported that HDACi (trichostatin A and VPA) triggers apoptosis in human leukemic cell lines expressing RUNX1‐related chimeric proteins such as RUNX1‐ETO, TEL‐RUNX1 or RUNX1‐EVI1.^{( 74 )} A cell line without RUNX1‐related chimeras is less affected by the HDACi treatment. These data suggest that HDACi seems to be an attractive choice in the molecular targeting therapy of RUNX1‐EVI1‐expressing leukemia.

Figure 3.

RUNX1‐EVI1 is a multifunctional oncoprotein. RUNX1‐EVI1 exerts oncogenic function through three pathways: dominant‐negative effect over RUNX1, repression of TGFβ signaling and inhibition of C/EBPα activity.

Another potential therapeutic agent targeted to RUNX1‐EVI1 is arsenic trioxide (ATO). Shackelford et al.^{( 75 )} found that ATO degrades RUNX1‐EVI1. The ATO treatment induces the differentiation and apoptosis of RUNX1‐EVI1‐expressing leukemic cells in vitro and elongates the survival of mice transplanted with these cells in vivo. They also demonstrated that ATO targets RUNX1‐EVI1 via two moieties, MDS1 and EVI1 moieties. The EVI1 induces a ubiquitin‐proteasome pathway and MDS1 induces a proteasome‐independent pathway. With abundant experiences of clinical use in treating acute promyelocytic leukemia, ATO could be used clinically as a targeted therapy for RUNX1‐EVI1‐positive human leukemia.

Summary

The fusion protein RUNX1‐EVI1 is a multifunctional protein demonstrated by its diverse role in regulating differentiation, proliferation, apoptosis and self‐renewal capacity. It functions as a dominant‐negative suppressor of RUNX1 and has oncogenic properties of EVI1. There is no doubt that RUNX1‐EVI1 plays a major role in t(3;21)‐related MDS and MDS/AML. It remains to be determined whether RUNX1‐EVI1 induces leukemia by itself, or needs additional genetic events. A conditional knock‐in experiment of the RUNX1‐EVI1 gene would help clarify this question.