Downregulation of RUNX1/CBFβ by MLL fusion proteins enhances hematopoietic stem cell self-renewal

The CBF (core binding factor) heterodimeric transcription factor complex, composed of RUNX1 and CBFβ subunits, has been shown to play critical roles in both embryogenesis and hematopoiesis.^1,-3  RUNX1 binds to its DNA target sequences through an evolutionarily conserved runt domain⁴  and can both activate and repress the transcription of its target genes. CBFβ does not bind to DNA directly. However, CBFβ induces a conformational change in RUNX1, which increases RUNX1 DNA-binding ability.^5,6  Dysregulation of RUNX1/CBFβ, through chromosome translocations, mutations, microdeletions, or transcriptional downregulation, is leukemogenic.⁷ 

The stability of RUNX1 is tightly regulated. We have demonstrated that CBFβ enhances RUNX1 stability by preventing its ubiquitin-mediated degradation in the proteasome.⁸  Furthermore, it has been shown that the CHIP/Stub1 E3 ubiquitin ligase can mediate RUNX1 ubiquitination and degradation independently of the molecular chaperones Hsp70/90.⁹  mSin3A interacts with unphosphorylated RUNX1 and protects it from proteasome-mediated degradation. Extracellular signal-regulated kinase–dependent RUNX1 phosphorylation induces the release of RUNX1 from mSin3A, resulting in degradation of RUNX1 in a time-dependent manner.^10,11  RUNX1 protein levels are also regulated during the cell cycle. RUNX1 protein is increased in S and G2/M phase cells compared with G1 phase cells, and this regulation is independent of cytokine-induced RUNX1 phosphorylation.¹² 

The mixed lineage leukemia (MLL) protein is a mammalian homolog of the trithorax proteins,^13,14  which are critical for hematopoiesis and the self-renewal of adult stem cells.¹⁵  Full-length MLL is a large multidomain protein of 3969 amino acid (aa) residues including 3 AT-hook DNA-binding motifs, a cysteine-rich CXXC domain with homology to DNA methyltransferases, 4 plant homeodomain (PHD) finger motifs, a PHD-flanking bromodomain, a C-terminal transactivation domain, and a Suvar3-9, enhancer-of-zeste, trithorax (SET) domain.^16,17  As with its Drosophila homolog, trx, MLL regulates target gene expression by methylating lysine 4 of histone H3 (H3K4) through the methyltransferase activity of its C-terminal SET domain; methylation of H3K4 is closely associated with transcriptional activation.¹⁸  MLL can be cleaved to an N-terminal 300-kDa fragment and a C-terminal 180-kDa fragment by taspase.¹⁹  The human MLL gene is located on chromosome 11q23 and is often involved in chromosome translocations with various partner chromosomes, generating MLL fusion proteins.^20,-22  More than 70 MLL fusion proteins have been documented in leukemia patients.^23,24  In almost all fusion proteins, MLL breaks within an 8.3-kb break point cluster region (BCR),²⁵  which results in the deletion of PHD finger region but also the maintenance of the MLL CXXC domain within the fusion protein. Interestingly, similar break points are also found in MLL partial tandem duplications (MLL-PTDs), which result from partial duplication within the 5′ region of the MLL gene. These duplications consist of an in-frame repetition of MLL exons in the 5′-3′ direction and produce an elongated protein.²⁶  The incidence of MLL-PTD was 6.4% in unselected adult and childhood acute myeloid leukemia (AML) and 5% in myelodysplastic syndromes.^27,28 

MLL regulates many targets involved in self-renewal, proliferation, survival, and differentiation.^22,29,30  The most well-studied targets are found in the HOXA gene cluster. MLL may bind to DNA or chromatin directly or be recruited to target loci by DNA-binding transcription factors.^18,31,32  Our recent study showed that MLL, RUNX1, and CBFβ interact and form a complex.³³  MLL interacts with the N terminus of RUNX1 (51-106 aa), and prevents RUNX1 from ubiquitin-mediated degradation. Although CBFβ does not interact with MLL directly, it can strongly enhance the interaction between RUNX1 and MLL. RUNX1/CBFβ recruits MLL to the regulatory regions of the PU.1 gene, which is important for maintaining the H3K4 trimethylation of the PU.1 upstream regulatory region and promoter regions.³⁴  However, the functional consequence of MLL fusions on RUNX1/CBFβ activity has not been fully understood.

In this study, we investigated the effects of MLL truncation mutants and its fusion proteins on RUNX1/CBFβ. We found that RUNX1 protein was not only downregulated by MLL fusion proteins, but also by MLL aas 1-1406, which are common to MLL fusion proteins). We confirmed this finding in Mll-Af9 knock-in mice and human M4/M5 MLL fusion–expressing AML cell lines. Using Runx1^+/−Cbfβ^+/− mice as a Runx1/Cbfβ hypomorph model, we found significant hematopoietic/stem progenitor cell (HSPC) expansion and higher repopulation activity. Overexpression of RUNX1 inhibits the development of AML in Mll-Af9 knock-in mice HSPCs. Conversely, reducing Runx1/Cbfb levels accelerates MLL-AF9–mediated AML in bone marrow transplantation (BMT) assays. Our data suggest that downregulation of RUNX1/CBFβ is critical for the development and maintenance of MLL translocation-related leukemia; therefore, targeting RUNX1/CBFβ levels may be a potential therapy for MLLs.

Methods and materials used in this study can be found in the supplemental data on the Blood Web site. All animal studies were conducted according to approved Institutional Animal Care and Use Committee protocol and federal regulations.

To understand the impact of MLL fusion protein expression on RUNX1 and CBFβ, either MLL, MLL-BP (1-1406), or MLL fusions were coexpressed with RUNX1, CBFβ, or both RUNX1 and CBFβ in 293T cells (Figure 1A). We found that MLL-BP and the 3 MLL fusion proteins all decreased RUNX1 levels, and MLL-eleven nineteen leukemia (ENL) caused a greater decrease in RUNX1 compared with MLL-AF9 and MLL-AF4 fusion proteins (Figure 1B and supplemental Figure 1A). CBFβ protein was mildly decreased by MLL-BP and MLL fusions when expressed alone (Figure 1C and supplemental Figure 1B); however, when CBFβ was coexpressed with RUNX1, it was significantly decreased, indicating that the full decrease in CBFβ by MLL-BP and MLL fusions depends on RUNX1 (Figure 1D and supplemental Figure 1C). We also coexpressed either GATA-1 or C/EBPα with MLL-BP. The level of each transcription factor remained unaltered by the coexpression of MLL-BP (supplemental Figure 2), which suggests that MLL-BP has specificity for RUNX1/CBFβ. To confirm this finding, we transduced retroviruses containing MLL-BP and MLL-AF9 into U937 cells and found that both of them, but not empty virus, downregulated RUNX1 and CBFβ proteins in U937 cells (Figure 1G and supplemental Figure 1D).

To validate this finding in human AML, various leukemia cell lines were collected and separated into 3 groups based on AML subtype. RUNX1 and CBFβ protein levels were higher in the cells lines that do not have MLL translocations (groups A and B) than in the cell lines with MLL translocations (group C) (Figure 1E and supplemental Figure 1E). The Kasumi-1 cell line, which contains only 1 wild-type RUNX1 allele, exhibited higher RUNX1 protein levels than the cells lines with MLL translocations and 2 RUNX1 alleles. However, expression levels of Menin, a transcription factor that interacts with the N terminus of MLL, did not differ among these groups of leukemia cell lines (Figure 1E and supplemental Figure 1E). We found no significant difference in the levels of RUNX1, CBFβ, and Menin messenger RNA (mRNA) between the groups of M4/M5 AMLs with or without MLL translocations (Figure 1F). We also analyzed RUNX1, CBFβ, and MEN1 (Menin) mRNA levels in published AML patient samples.³⁵  Meta-analysis showed that RUNX1, CBFβ, and MEN1 mRNA levels are not decreased in M4/M5 subtypes with MLL translocations compared with the same subtype without MLL translocations, consistent with our AML cell line data (supplemental Table 2 and supplemental Figure 3A,C).

We further tested whether MLL fusion proteins could downregulate endogenous RUNX1 and CBFβ levels in human cord blood CD34⁺ cells using a Tet-off inducible expression system.³⁶  Forty-eight hours after adding doxycycline, MLL-AF9 protein levels decreased, whereas RUNX1 protein levels increased. However, neither RUNX1 nor CBFβ protein levels were changed by the presence of doxycycline in CD34⁺-derived cell line with stable expression of MLL-AF9 (Figure 1H).

To examine the effects of MLL-BP and MLL fusions on endogenous Runx1 and Cbfβ expression in murine cells, bone marrow (BM) cells from wild-type C57BL/6 mice were transduced with retroviruses containing either Meis1, MLL-BP, or MLL-AF9 and plated for colony-forming unit (CFU) assays. BM cells transduced with MLL-BP virus had enhanced replating potential relative to those transduced with empty or Meis1 virus in the third plating, but there were no colonies in the fourth plating. Only MLL-AF9–transduced cells had replating ability, indicating a gain-of-function activity of MLL-AF9 (Figure 1I). We confirmed that Runx1 and Cbfβ were both downregulated in MLL-BP– and MLL-AF9–expressing cells, but not in Meis1-expressing cells (Figure 1J). Our data indicate that Runx1/Cbfβ downregulation is MLL-BP dependent but is independent of the MLL downstream target, Meis1.

We have shown that wild-type MLL can stabilize the RUNX1/CBFβ complex.³³  In MLL fusion leukemias, only 1 allele of MLL is truncated and fused with another gene, whereas the other allele is intact. A dominant-negative function of the MLL fusions over the remaining wild-type allele has been suggested; thus, we researched whether MLL fusion protein could downregulate RUNX1/CBFβ in the presence of wild-type MLL. We cotransfected RUNX1/CBFβ with different MLLs:MLL-AF9 ratios and found that MLL-AF9 has a strong dominant-negative effect over MLL, even at high MLL:MLL-AF9 ratios (Figure 1K).

We used cycloheximide to determine whether MLL fusion proteins or MLL-BP alters the half-life of RUNX1 protein. RUNX1 had a shorter half-life in the presence of MLL-BP and an even shorter half-life in the presence of MLL-ENL, compared with the vector control (Figure 2A-B). Conversely, we noted that RUNX1 half-life was prolonged by full-length MLL (Figure 2A-B), consistent with our previous findings.³³ 

Having demonstrated that RUNX1 can be degraded through the proteasome pathway,⁸  we tested whether MLL-BP and MLL fusions downregulate RUNX1 protein stability via the proteasome pathway. We found that MG132 can only partially rescue RUNX1/CBFβ from downregulation by MLL-BP and MLL fusion proteins (Figure 2C). We also assessed RUNX1/CBFβ protein levels in the MV4-11 (MLL-AF4⁺) and SKM1 (no MLL translocation) cell lines using MG132 or another proteasome inhibitor, lactacystin. Both proteasome inhibitors increased endogenous RUNX1 and CBFβ protein levels in these 2 cell lines, suggesting that proteasome inhibitors effectively stabilize RUNX1/CBFβ, regardless of the presence of MLL fusion proteins (Figure 2D). In contrast, several common protease inhibitors did not rescue RUNX1 protein degradation in the MLL fusion expressing the human leukemia cell lines MV4-11 (MLL-AF4⁺) and THP1 (MLL-AF9⁺) (supplemental Figure 4). Having shown that MLL decreases the poly-ubiquitination of RUNX1,³³  we tested whether MLL-BP or MLL fusion proteins could alter RUNX1 ubiquitination; MLL-BP, MLL-AF9, and MLL-ENL all led to increased poly-ubiquitination of RUNX1 (Figure 2E and supplemental Figure 1F).

Because lysine residues are the target sites for ubiquitination, we generated a mutant RUNX1 construct in which all 9 lysine residues were mutated to arginine residues, RUNX1 (9K/R) (Figure 2F), and examined how RUNX1 (9K/R) protein stability was affected by MLL fusion proteins. As expected, RUNX1 (9K/R) was expressed at higher levels than wild-type RUNX1 (Figure 2G and data not shown). However, RUNX1 (9K/R) could still be stabilized by full-length MLL and be degraded by MLL-BP and MLL-ENL. To rule out the possibility that ubiquitin molecules were attaching to residues other than lysine on RUNX1 (for example, the N-terminal aa), we used MG132 to treat 293T expressing either wild-type RUNX1 or RUNX1 (9K/R). Wild-type RUNX1 protein levels increased when treated with MG132, whereas RUNX1 (9K/R) protein levels were minimally changed in the presence of MG132 (Figure 2H). mRNA expression levels for all samples were unchanged upon MG132 treatment (data not shown). These data provide further evidence that the ubiquitin pathway, although involved in MLL-BP and MLL fusion protein–mediated RUNX1/CBFβ downregulation to some degree, is unlikely to be the major mechanism by which MLL-BP or MLL fusion proteins downregulate RUNX1/CBFβ protein.

The MLL-BP is the common region for all MLL fusions and is duplicated in MLL-PTD. To address which region of MLL-BP is responsible for downregulation of RUNX1, we constructed a series of MLL deletion mutants (Figure 3A). We found that the MLL constructs—1-1406, 1-1362, 1-1211, and 1-1194—downregulated RUNX1 protein, whereas MLL deletion mutants that lacked the CXXC domain had almost no effect on RUNX1 stability (Figure 3B). This finding strongly suggests that the CXXC domain is required for RUNX1 downregulation. We generated expression constructs in which either the CXXC domain was deleted or contained point mutations that destroy the structure of CXXC zinc finger (C1155A, C1158A).^37,38  Each of these constructs lacked the ability to decrease RUNX1 protein levels (Figure 3C), whereas a CXXC mutant of MLL-BP that maintains the CXXC zinc finger structure (E1165A³⁷ ) retained the ability to downregulate RUNX1 (Figure 3C). We also cloned the CXXC domain and flanking regions from MLL1, MLL2/KMT2B, and CGBP (Figure 3A) to determine whether overexpression of only the CXXC domain and flanking regions could downregulate RUNX1/CBFβ. We found that minimal domains from both MLL1 and MLL2/KMT2B downregulate RUNX1/CBFβ significantly, whereas similar domains from CGBP had no effect (Figure 3D).

To extend our findings in vivo, LSK (Lin⁻ c-kit⁺ sca1⁺) populations were isolated from the BM of wild-type or Mll-Af9 knock-in mice and the protein and mRNA expression levels of Runx1 and Cbfβ were determined. We found Runx1/Cbfβ protein levels were lower in Mll-Af9 knock-in LSK vs wild-type LSK (Figure 4A-B) with no significant change in Runx1/Cbfβ mRNA (Figure 4C), consistent with what we found in human AML cell lines.

As expected, Mll-Af9 knock-in BM cells or splenocytes can replate in vitro and rapidly develop AML in vivo. In contrast, BM or spleen cells from polyinosinic:polycytidylic acid injected Runx1^flox/floxMx1-Cre⁺ (Runx1^Δ/Δ) mice show enhanced replating activity in vitro, but do not develop spontaneous AML.³⁴  To determine whether intermediate levels of Runx1/Cbfβ expression impact colony-replating activity, BM cells from Runx1^+/−Cbfβ^+/− mice were analyzed in CFU assays (Figure 4D-E). In fact, Runx1^+/−Cbfβ^+/− BM cells produced significantly more colonies in the second and third plating compared with wild-type BM cells. Runx1^Δ/Δ and Mll-Af9 knock-in BM cells also produced significantly more colonies than wild-type BM cells. These data indicate that Runx1^+/−Cbfβ^+/− BM cells can maintain an immature phenotype more easily than wild-type BM cells.

We have demonstrated that MLL translocations result in decreased levels of RUNX1/CBFβ, a hypomorph condition; however, the significance of reduced RUNX1/CBFβ levels has not been examined thus far. We generated Runx1^+/−Cbfβ^+/− compound heterozygous mice as a Runx1/Cbfβ hypomorph model (Figure 5A). We analyzed hematopoiesis in Runx1^+/−Cbfβ^+/− mice with a specific focus on the function of HSPC as previously described.³⁹  BM cells from Runx1^+/−Cbfβ^+/− mice exhibited a higher percentage of LSK cells in the Lin⁻ population as compared with LSK from wild-type mice (Figure 5B-C) and statistically fewer CMP within the hematopoietic cell (HPC, Lin⁻ c-kit⁺ sca1⁺) population (P < .01, Figure 5D). The percentage of Lin⁻ c-kit⁺ sca1⁻ (LK, HPC) cells in the Lin⁻ population as well as the Lin⁻c-Kit⁺Sca1⁻CD34⁺CD16/32⁺ and (Lin⁻c-Kit⁺Sca1⁻CD34⁻CD16/32⁻ percentages in the HPC populations (Figure 5C-D) were comparable between the 2 genotypes.

We also performed standard CFU-spleen assays to dissect the function of Runx1^+/−Cbfβ^+/− HSPCs. On days 8 and 12 of a CFU-spleen assay, we found a 39% and 30% increase in CFUs from Runx1^+/−Cbfβ^+/− BM compared with wild-type BM, respectively (P < .01, Figure 5E). These data indicate hematopoietic stem cell (HSC) expansion within the Runx1^+/−Cbfβ^+/− BM. We further analyzed the in vivo reconstituting ability of BM cells obtained from Runx1^+/−Cbfβ^+/− mice. We found these BM cells have greater engraftment potential and long-term reconstitution ability than the control wild-type BM cells in both first and second BMT assays (Figure 5F). These data show that Runx1^+/−Cbfβ^+/− mice have a significant expansion of phenotypic and functional HSPCs compared with their wild-type littermates. Thus, downregulation of Runx1/Cbfβ could contribute to MLL fusion leukemogenesis not only through blocking terminal differentiation, but also through expanding self-renewal within the HSPC pool, potentiating further accumulation of cooperative mutations and progression to full-blown leukemias.

Given that MLLs have low RUNX1 levels, to understand whether restoration of RUNX1 can impair the MLL phenotype, we overexpressed RUNX1 in MV4-11 cells and found cell growth arrest and morphological differentiation of sorted stable transfectants (Figure 6A-C). To validate this finding in an in vivo BMT assay, we used Mll-Af9 knock-in mice and overexpressed RUNX1 in BM cells from preleukemic (Figure 6D-F) or leukemic (Figure 6G-I) mice. Under both conditions, overexpression of RUNX1 resulted in terminal differentiation (Figure 6D,G) and reduced colony-forming ability (Figure 6E,H). More importantly, RUNX1 overexpression in Mll-Af9 knock-in HSPCs completely blocked their leukemic potential in BMT assays (Figure 6F,I).

Although MLLs have low RUNX1/CBFβ levels, further reducing the level by Runx1 deletion in an MLL-ENL model accelerated the leukemia.⁴⁰  However, residual RUNX1/CBFβ levels are absolutely required for MLL because complete deletion of both Runx1/Cbfβ blocked MLL formation.⁴¹  To understand the gene dosage effect in the context of MLL, we also took advantage of BM cells from the Runx1^flox/wt/Cbfβ^flox/wt/Rosa26-Cre-ERT2 mouse model. HSPCs were transduced with MLL-AF9, sorted for eGFP⁺ cells, and plated in the presence or absence of 4-hydroxytamoxifen (4-OHT) to induce hypomorphic Runx1/Cbfβ function. Deletion of 1 allele of Runx1 and Cbfβ resulted in significantly more colonies upon replating of MLL-AF9 cells (Figure 6J-K). We also performed BMT assays with the sorted eGFP⁺ cells. After confirming eGFP⁺ cell reconstitution in recipient mice, we injected 1 group of mice with tamoxifen and the other with corn oil, and found accelerated AML development in the tamoxifen-injected group (Figure 6L). We confirmed Runx1/Cbfβ gene deletion and protein level changes (supplemental Figures 5 and 6M).

Our study suggests a working model (Figure 7) for wild-type MLL cells, in which the downregulation of RUNX1/CBFβ proteins mediated by the CXXC domain is inhibited by the presence of the PHD finger domains; this results in a normal level of RUNX1/CBFβ that then allows HSC differentiation (Figure 7A). In the presence of MLL fusion proteins or MLL-PTD (Figure 7B), the loss of MLL-AFs or duplication of the CXXC domain (MLL-PTD) leads to downregulation of RUNX1/CBFβ, which promotes self-renewal and differentiation blockade. Additional mutations can then cooperate to cause leukemia.

The leukemogenic 11q23 translocations fuse the MLL N terminus with a wide variety of fusion partners, ranging from nuclear factors to cytoplasmic proteins.^16,29,30  More than 70 MLL fusion partners have been identified, and the number is increasing. Many previous studies on MLL fusion proteins have focused on the loss of the MLL C terminus and the “gain of function” provided by the different fusion partners. Our study reveals an unreported function of MLL fusion proteins, mediated by a common protein domain within the NH2 terminus of MLL, that downregulates RUNX1/CBFβ. This property contrasts with our previous demonstration that full-length MLL prevents ubiquitin-proteasome–mediated degradation of RUNX1 by reducing poly-ubiquitination of RUNX1.³³ 

All MLL translocations occur in the BCR of MLL that spans exons 9-11, indicating that all MLL fusion proteins retain the CXXC domain and the adjacent RD2 region while deleting the PHD finger domain and the MLL C terminus including the SET domain. MLL-PTD is another mutant form of MLL, which contains a partial tandem duplication of the MLL N terminus, with the duplicated breakpoints matching the BCR found in MLL translocations (exons 9-11).^22,27 

Our study indicates that the CXXC domain is the part of the MLL N terminus responsible for the downregulation of RUNX1/CBFβ. The CXXC domain of MLL is a cysteine-rich sequence containing 2 CGXCXXC repeats in its core region (1154-1194 aa), which adopts an extended crescent-like structure coordinating 2 zinc ions.^37,42  The CXXC domain of MLL plays a crucial role in myeloid cell transformation, which correlates with its recognition of nonmethylated CpG dinucleotides.^37,38  Although the carboxyl-flanking region of the MLL CXXC domain (the “post-CxxC” moiety of MLL, which is rich in basic aa) is not required for binding to CpG sites,³⁷  it also contributes to MLL-associated myeloid transformation.^37,43  We found that the MLL deletion construct lacking the CXXC carboxy-flanking region (MLL 1-1194, Figure 3B) could downregulate RUNX1, although not as efficiently as the MLL deletion construct that retains both the core CXXC motif and the post-CXXC region (MLL 1-1211, Figure 3B), suggesting that the CXXC flanking region may play a synergistic role with the core CXXC motif in RUNX1 downregulation either through stabilization of the structure or through a protein–protein interaction.

Recently, the polymerase-associated factor complex (PAFc) has been found to interact with the CXXC-RD2 region in MLL.⁴⁴  PAFc stimulates MLL and MLL fusion protein—mediated transcriptional activation of HOXA9; disrupting the MLL-PAFc interaction selectively inhibits the growth of MLL-fusion leukemic cells.⁴⁵  Whether PAFc and its associated complexes and factors (eg, super elongation complex,⁴⁶  BRE1/RAD6 complex⁴⁷ ) are also involved in regulating RUNX1/CBFβ will be important to dissect in future studies.

The MLL PHD finger domains region (MLL 1406-2666, shown in Figure 3A as the “inhibitory region” of RUNX1 downregulation) is lost in all MLL fusion proteins. Insertion of PHD finger domains into MLL-ENL or MLL-AF9 suppresses MLL fusion protein–induced HSC immortalization, indicating that the loss of the PHD finger domains in MLL fusion proteins is necessary for leukemogenesis.^48,49  In our study, the 300-kDa fragment of MLL generated by taspase cleavage (MLL 1-2666), which retains the CXXC domain and all the PHD finger domains, showed no effect on RUNX1 downregulation (Figure 3B). Thus, the mechanism by which the PHD finger domains block the CXXC effects on RUNX1/CBFβ will be an interesting question for future study.

The downregulation of CBF (RUNX1/CBFβ) is involved in leukemogenesis. It has been shown that depletion of Aml1/Runx1 accelerates the development of MLL-ENL in a mouse BMT model.⁴⁰  We found that introduction of MLL-BP into wild-type BM cells conferred better replating potential (Figure 1I) and that the hypomorph condition of Runx1/Cbfβ in Runx1^+/−Cbfβ^+/− mice resulted in HSPC expansion (Figure 5). Conversely, RUNX1/CBFβ upregulation could be a potential therapeutic target for MLLs as an alternative therapy approach.

In our study, overexpression of RUNX1 inhibited the AML development of Mll-Af9 knock-in HSPCs (Figure 6A-I); consistent with this, further reducing of Runx1/Cbfβ levels accelerated Mll-Af9–mediated AML in BMT assays (Figure 6J-M). These results are fully consistent with the tumor suppressor role of Runx1/Cbfβ. At the same time, we recently showed that a certain level of RUNX1 is required for the growth and survival of MLL-AF9 cells,⁴¹  and a recent report showed that RUNX1 expression is important for the growth of t(4;11) leukemia cell lines, in which it plays a role in the activation of specific target genes by direct interaction with AF4-MLL.^21,50  Taken together, these data argue that RUNX1 level is tightly controlled and low RUNX1/CBFβ activity is critical for leukemogenesis. Therefore, modulating RUNX1/CBFβ dosage, either by overexpression or by further inhibition, could offer a therapeutic strategy for the treatment of MLL.

In summary, we have demonstrated a new mechanism of leukemogenesis mediated by MLL translocations. The common portion of the MLL fusion proteins (the approximately 1400-aa region of the MLL N terminus) can downregulate RUNX1/CBFβ protein; this function is dependent on its CXXC domain. Losing the PHD finger domains exposes this negative regulatory activity. RUNX1/CBFβ hypomorphic mice have HSPC expansion, suggesting that decreased RUNX1/CBFβ dosage could potentially contribute to the initiation and maintenance of AML. Modulating RUNX1 protein levels or RUNX1/CBFβ activity, either up or down, is a new therapeutic strategy for leukemias bearing rearrangements of chromosome 11q23.

The authors thank Dr H. Leighton Grimes and Ashish R. Kumar for careful reading and giving of advice on experiments and the manuscript.

This work was supported by grants from Ohio Cancer Research Associates, Cancer Free Kids, Leukemia Research Foundation and Pilot Research Grant of the State Key Laboratory of Experimental Hematology (Tianjin, China) (G.H.), Leukemia Lymphoma Society SCOR grant (S.D.N.), and the National Natural Science Foundation of China grants 81200356 (X.Z.) and 81000220 (F.H.).

Contribution: X.Z., X.Y., and G.H. designed the research; X.Z., A.C., X.Y., Y.Z., F.H., Y.H., Y.D., Y.R., B.L., R.M.C., S.E.E., N.H., and G.H. performed research; X.Z., Y.Z., F.H., and G.H. analyzed data; A.M.-D., J.Z., Z.X., W.T., D.G.T., Q.W., W.C., J.C.M., and S.D.N. contributed vital new reagents; and X.Z. and G.H. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Gang Huang, Cincinnati Children's Hospital Medical Center, 3333 Burnet Ave, Room S7.607, MLC 7013, Cincinnati, OH 45229-3039; e-mail: gang.huang@cchmc.org.