The Pathogenesis of Mixed Lineage Leukemia

Abstract

Aggressive leukemias arise in both children and adults as a result of rearrangements to the Mixed Lineage Leukemia (MLL) gene located on chromosome 11q23. The MLL gene encodes a large histone methyltransferase that directly binds and positively regulates gene transcription including HOX genes. MLL is involved in chromosomal translocations, partial tandem duplication and amplifications, all of which result in hematopoietic malignancies due to sustained HOX expression and stalled differentiation. MLL lesions are associated with both acute myeloid leukemia (AML) and acute lymphoid leukemia (ALL) and are usually associated with a relatively poor prognosis despite improved treatment options like allogeneic hematopoietic stem cell transplantation underscoring the need for new treatment regimens. Recent advances have begun to reveal the molecular mechanisms driving MLL associated leukemias which have provided opportunities for therapeutic development. Here we discuss the etiology of MLL leukemias and potential directions for therapeutic development.

Introduction

Epigenetic modifications of chromatin constitute a form of cellular memory that allows gene expression programs to remain intact through cell division. Transcriptional reprogramming and change in epigenetic signatures due to alterations in transcription factors can have catastrophic effects on cellular development including oncogenic transformation. In this review we focus on the clinical characteristics and mechanisms of transformation of leukemias associated with rearrangement of the epigenetic modifier MLL.

1. Cytogenetics and clinical characteristics of MLL associated leukemias

Chromosomal rearrangements at 11q23 are associated with pediatric, adult and therapy-related leukemias and led to the discovery of the Mixed Lineage Leukemia (MLL) gene (1-3). MLL rearrangements fuse the N-terminus of MLL to a fusion partner protein (Figure 2) and constitute >70% of infant acute lymphoid leukemias (ALL) and between 35-50% of infant acute myeloid leukemia (AML); MLL translocations also occur in leukemias of older children and adults, overall accounting for 10% of cases (4). Secondary or therapy-related leukemias that arise in patients treated with topoisomerase II inhibitors for other malignancies also display MLL rearrangements and account for about 5-10% of MLL associated leukemias (4). Together, about 10% of human acute leukemias harbor MLL translocations (4). MLL is also affected by abnormalities other than chromosomal translocations including partial tandem duplications (PTDs) of exons encoding the N-terminus of MLL in about 10% of AML (Figure 2). Non-rearranged MLL is also occasionally amplified in myelodysplasia and AML (5). In general, patients with MLL rearrangements have a poor prognosis and are treated according to high-risk protocols, however this can vary depending on the translocation partner (6). MLL is involved in over 100 recurrent translocations and greater than 60 different partner genes have been identified (6). Despite the vast number of partner genes, MLL is predominantly recombined to nine translocation partners accounting for almost 90% of MLL rearrangements (Figure 1) (6). These include AF4, AF9, ENL, AF10, AF6, ELL, AF1P, AF17 and SEPT6 (Figure 1).

Figure 2.

Structure of wild type and leukemia associated MLL proteins. Top: Domain architecture of wild type MLL. MLL is a large multi-domain protein of about 4000 amino acids. Cleavage of MLL (denoted by the blue arrow) results in 320 kDa MLL^N and 180 kDa MLL^C fragments that non-covalently associate. Domains within MLL^N include three AT hooks (red), two subnuclear localization motifs (SNL) (purple), a DNMT1 homology region (CxxC) (Green), four plant homeodomain (PHD) fingers (blue), an atypical bromodomain (orange) and a FYRN domain (open circle). The Breakpoint Cluster Region (BCR) spans an 8.3 kb region bound by BamH1 restriction sites and encompasses exons 5-11 or 7-13 using old or new nomenclature respectively and is the site of chromosomal translocations involving MLL. Between the CxxC and first PHD finger is repression domain 2 (RD2) that is rich in basic amino acids. A HCF binding motif (HBM) (yellow) is found between the bromodomain and PHD3. MLL^C contains a transactivation domain (TAD) (filled oval), a FYRC domain (open square) and C-terminal SET domain (dark blue). Middle: Chromosomal translocations involving MLL result in chimeric MLL fusion proteins that include N-terminal sequence of MLL up to the BCR (dotted vertical line) followed by one of several different fusion partners. Examples of fusion partner proteins including AF4, ENL, AF9 and AF6 are shown. MLL fusion proteins invariably retain AT-hooks, SNL1/2 and the CxxC domain of MLL^N while losing the downstream PHD fingers and beyond. Bottom: The MLL gene is also prone to internal tandem duplications (MLL-PTD) resulting in duplication of MLL sequences comprising the AT-hooks, SNL1/2 and the CxxC domain which are inserted at the BCR.

Figure 1.

Translocations with MLL occur with a large and diverse group of partner genes. The frequency of some of the most common translocations with MLL are shown for both pediatric and adult ALL or AML. In general, MLL is fused with a more diverse group of partner proteins in AML compared to ALL, which is primarily composed of AF4, AF9 and ENL translocations (6).

In both pediatric and adult ALL cases involving MLL, the most common translocation partner is AF4 (AFF1) followed by AF9 and ENL (Figure 1). The t(4;11) translocation results in the MLL-AF4 fusion protein and is mostly associated with cases of CD19+ B-lineage ALL. Infants (under the age of 1) with t(4;11) have a particularly poor 5-year event free survival (EFS) rate of about 19%, which increases to about 42% in children over the age of 1 (7, 8). This is striking given a 74-96% EFS in infants with ALL lacking MLL rearrangements (9). Although great strides have been made in infant ALL treatment, including allogeneic hematopoietic stem cell transplantation, transplantation results in worsened EFS compared to chemotherapy alone (8).

Patients with AML have a more broad distribution of translocation partners compared with the ALL, with the most common being the AF9 and ENL translocations. The t(9;11) arrangement produces the MLL-AF9 fusion protein which is found in about 2-5% of all AML and up to 25% of de novo AML in children (7, 10). The prognosis for patients carrying the t(9;11) is more favorable than other 11q23 rearrangements, however the median survival for de novo cases is only ~4 years (10, 11). The MLL-ENL fusion protein associated with the t(11;19) rearrangement is associated with both AML and ALL. Most MLL-ENL cases are found in infants of less than 1 year of age with a bi-phenotypic or B-cell ALL. The median survival for these patients is very poor with a median survival of < 1 year (8). The differences in disease phenotype and patient survival rates based on translocation partner and MLL lesion suggests a better understanding of the molecular events surrounding these various translocation events may translate to better treatment of MLL-associated leukemias.

2. The functions and structure of wild type MLL

MLL (3969 amino acids) is the mammalian homolog of the trithorax (trx) protein found in Drosophila. MLL is a member of the evolutionarily-conserved trithorax group (trxG) family of proteins that positively regulate gene transcription and act antagonistically to the Polycomb group (PcG) proteins (12). The trxG family was first identified through studies of mutant flies, which show homeotic transformations (phenotype where body segment is transformed into a different segment) due to improper expression of the homeobox (Hox) genes (13). These defects were rescued by mutations to PcG genes. It was latter shown that the PcG proteins function as transcriptional repressors of the same targets as trxG proteins, namely the Hox genes. Hox genes are transcriptional regulators that are integral to the formation of the body plan during embryogenesis. Hox genes are regulated in a strict spatiotemporal manner that is critical for tissue development including the hematopoietic system (14). Recent work has demonstrated that many developmentally regulated genes, including Hox genes, display “bivalent” epigenetic signatures, including H3K4 methylation (associated with active transcription) and K3K27 methylation (associated with a transcriptionally repressed locus). These epigenetic marks are deposited, in part, by the MLL complex and the EZH2 containing Polycomb Repressive Complex 2 (PRC2) respectively and allow the target genes to be “poised” for transcriptional activation by RNA polymerase II (15-17).

MLL is expressed in most tissues, including myeloid and lymphoid cells, and positively regulates expression of the clustered Hox genes through histone H3 lysine 4 (H3K4) methyltransferase activity (18, 19). Knockout studies in mice show that Mll is important for the maintenance of Hox gene expression. Deletion of Mll leads to embryonic lethality at ~E10.5 indicating that Mll is essential for proper development. In keeping with its role as a trxG protein disruptions of Mll result in homeotic transformations in Mll knockout mice including posterior shifts in Hox expression patterns as well as defects of the axial skeleton and the hematopoietic system (20, 21). The expression of Hox genes is initiated normally in Mll knockout mice but is not properly maintained. Detailed analysis of the hematopoietic system revealed that Mll is necessary for the proliferation and/or survival of the hematopoietic stem cell (HSC) and progenitor compartment in both the developing fetus and adult mice (20-24).

MLL is proteolytically cleaved by the threonine-aspartase TASPASE1 into a larger 320 kDa N-terminal fragment (MLL^N) and smaller 180 kDa C-terminal fragment (MLL^C) (Figure 2) (25, 26). Similar to Mll-/- mice, Taspase1-/- mice display homeotic defects due to improper Hox expression (27). After cleavage the two fragments non-covalently associate, via interaction between the FYRN and FYRC domains on MLL^N and MLL^C respectively, and translocate into the nucleus (Figure 2). Two subnuclear localization signals in MLL^N localize MLL to subnuclear punctuate spots. MLL^N contains several functional domains involved in binding DNA including three N-terminal AT-hooks, which nonspecifically bind the minor groove of DNA and a DNA methyltransferase homology region (or CxxC domain) that specifically binds unmethylated DNA (Figure 2) (5, 28-31). This region, along with the lysine rich RD2 region, which lies immediately downstream of the CxxC domain, has been reported to have inherent transcriptional repression activity (Figure 2) (32). A group of four plant homeodomain (PHD) zinc fingers with an embedded bromodomain are also present within MLL^N. PHD finger 3 binds to tri-methylated H3K4, which also aids in MLL recruitment to target loci (33, 34). Little is understood about the other PHD fingers. Although bromodomains on other epigenetic regulators (like GCN5) bind strongly to acetylated histones, the bromodomain on MLL does not appear to have this affinity (35). Within MLL^C lies a transcriptional activation domain that recruits the histone acetyltransferase CREB-binding protein (CBP) (36). At the C-terminus of MLL^C is a SET (Su(var)3-9, enhancer of zeste, trithorax) domain which is responsible for the H3K4 methyltransferase activity of MLL (Figure 2) (18, 19).

3. The interacting proteins and molecular biology of MLL and MLL fusion proteins

MLL functions within the context of a large multi-protein complex including MLL, WDR5, RbBP5 and ASH2L, which is required for maximal enzymatic activity. This core complex is shared with other H3K4 methyltransferases including SET1, MLL2 and MLL3. Each component is required for full activity of the complex (37, 38). Detailed analysis demonstrated that WDR5 mediates interaction between the MLL^C catalytic unit and the core complex as well as the histone H3 substrate (39). In addition to the core complex, MLL^C associates with the histone H4K16 acetyltransferase MOF. Full transcriptional activation of the MLL target gene Hoxa9 requires both MLL associated H3K4 methylation and MOF associated H4K16 acetylation and explains the correlated distribution of these marks on active genes (40). As mentioned above, the CBP/p300 histone acetyltransferase also associate with MLL^C and contributes to transcriptional activation (Figure 3A) (36).

Figure 3.

MLL interacting proteins. A) MLL is known to interact with a variety of partner proteins. Sequences at the extreme N-terminus of MLL are necessary for interaction with Menin and LEDGF. Together these proteins form a trimeric complex that is necessary for leukemogenesis. PAFc interacts with sequences of MLL retained in MLL fusion proteins and is also critically required for MLL fusion leukemogenesis. The RD2 region of MLL interacts with the co-repressors CtBP, the PcG proteins HPC2 and BMI-1, and the histone deacetylase HDAC1 and appears to be regulated by the binding of the cyclophilin Cyp33 to the third PHD finger of MLL. Host Cell Factor 1 and 2 (HCF1/2) interacts with a HBM consensus sequence found between the bromodomain and PHD4 and links MLL to the function of E2F proteins. The transactivation domain of MLL recruits the HAT CBP, which promotes histone acetylation and gene transcription. MOF associates with MLL^C and delivers H4K16 acetyltransferase activity to MLL target genes. MLL^C also associates with a core complex of proteins including RbBP5, WDR5 and ASH2L, which are necessary for MLL methyltransferase activity. B) Two Menin Binding Motifs (MBM) with varied binding affinities are found at the N-terminus of MLL and necessary for proper interactions with Menin. C) MLL associates with the PAF complex through two interaction sites. The crystal structure of the MLL CxxC domain shows how the pre and post CxxC domain are in close proximity because of the hairpin folding of the CxxC domain. This likely creates a single binding surface of MLL for making direct interactions with the PAF1 and CTR9 components of PAFc.

Although MLL functions as a transcriptional activator, MLL contains a repression domain adjacent to the CxxC domain that interacts with the co-repressors CtBP, the PcG proteins HPC2 and BMI-1, and the histone deacetylase HDAC1 (41). The binding of these repressive proteins is mediated by the binding of Cyp33, an RNA-binding nuclear cyclophilin with peptidyl-prolyl isomerase activity. The third PHD finger of MLL mediates the binding of Cyp33 that isomerizes a proline in the PHD3-Bromodomain linker region. The isomerization facilitates binding to Cyp33 which in turn mediates the recruitment of HDAC1 (35). C-terminal to the PHD3 and bromodomain is a HBM consensus sequence that is found in proteins that associate with the G1 phase regulator Host Cell Factor (HCF) proteins 1 and 2. HCF1/2 associates with both activator and repressor E2F proteins in a cell cycle-dependent manner. HCF1/2, in turn, recruits both SET1 and MLL HMT's to induce histone methylation and transcriptional activation (Figure 3A) (42, 43).

The protein-protein interactions described thus far are largely dependent on MLL sequences that are invariably lost in MLL-fusion proteins and consequently do not contribute to MLL-fusion protein transactivation. Although the AT-hooks and the CxxC domain are retained in MLL fusion proteins and are likely involved in DNA binding, these domains do not provide sequence specific targeting. Thus, alternative interactions must be in play for the proper function and targeting of MLL fusion proteins. At the extreme N-terminus of MLL is a Menin Binding Motif (MBM) (Figure 3A and B). The tumor suppressor protein Menin is coded by the MEN1 gene that is mutated in multiple endocrine neoplasia type 1. The N-terminal 43 amino acids of MLL contain two MBM sequences that mediate the interaction with Menin. Formation of a tri-molecular complex with both MLL and the chromatin-associated protein Lens Epithelium-Derived Growth Factor (LEDGF) is critical for the proper targeting of MLL or MLL fusion proteins to specific target genes, such as HOXA9, and for transcription upregulation required for leukemogenesis (44-47). Another interaction that is preserved in MLL fusion proteins and necessary for leukemogenesis is with the Polymerase Associated Factor complex (PAFc). This interaction is mediated between sequences before and after the CxxC domain that are consistently retained in MLL fusion proteins (Figure 3A and C) (34, 48). PAFc is a transcriptional activation complex that associates with RNA polymerase II and which facilitates the deposition of a mono-ubiquitination mark on histone H2B lysine 120, a pre-requisite for both H3K4 and K3K79 methylation (49-52). Like Menin, PAFc is required for MLL and MLL fusion protein recruitment to target loci. Thus it appears that MLL fusion proteins employ both protein-DNA interactions involving the AT-hooks and the CxxC domain and protein-protein interactions with Menin and PAFc for proper targeting to gene loci.

4. MLL target genes

The best understood targets of MLL and MLL fusion proteins are the HOX genes. As discussed, MLL and MLL fusions bind directly to the promoters and coding regions of HOX loci via DNA and PAFc and Menin/LEDGF interactions (18, 19, 53). The central role of HOX genes in MLL leukemogenesis is well documented and described in detail below. Another important direct target of MLL fusion proteins is EVI-1 (Ecotropic Viral Integration site-1), a nuclear transcription factor with essential roles in hematopoietic stem cell regulation. EVI-1 is upregulated in up to 10% of AML cases as a result of MLL translocation or by other mechanisms and is usually associated with a poor outcome (54). MLL is also known to regulate the cell cycle regulatory genes Cyclins and Cyclin Dependent Kinase Inhibitors (CDKIs). This is consistent with an MLL-E2F axis mediated by HCF-1 that targets MLL to S phase promoters, including Cyclins, which are regulated by E2F proteins (27, 43). The CDK inhibitors p27 and p18 are both regulated by MLL in a Menin-dependent manner (55, 56). The failed induction of these CDKIs as a result of inactivating Menin mutations is thought to be an important oncogenic pathway in these endocrine tumors (55). It has been difficult to reconcile how MLL is involved in regulating both cell cycle activators (Cyclins) and inhibitors (CDKIs). The observed temporal regulation of Cyclins by cell cycle specific recruitment of MLL may be due to a bi-phasic degradation of MLL protein through the ubiquitin-proteasome system (UPS) (57). The MLL protein is degraded during S phase and M phase by the SCF^SKP2 and APC^CDC20 E3 ubiquitin ligase complexes. Indeed, it was recognized in earlier studies that after proteolytic cleavage, the MLL^N fragment was extremely unstable when not in complex with MLL^C and that the MLL N-terminus may be stabilized in MLL fusion proteins (25, 58). Failure to properly degrade MLL at these cell cycle checkpoints results in defective cell cycle checkpoint response (59). Although ubiquitination of MLL is mediated through the N-terminus, MLL fusion proteins display resistance to degradation and thus may be a contributing mechanism in leukemogenesis.

Telomerase activity is upregulated in up to 90% of cancers and is instrumental in protecting chromosomes from telomere shortening. Recent work has established that MLL directs H3K4 methyltransferase activity to telomeres and induces transcription of telomere repeat-containing RNAs (60). Further, MLL fusion proteins have been shown to influence the transcription of a major subunit of the telomerase enzyme TERT, apparently through upregulation of HOXA7, which binds directly to the TERT promoter to induce expression (61).

5. Transcriptional pathways deregulated by MLL fusion proteins

MLL rearranged acute lymphoid and myeloid leukemias show highly characteristic gene expression patterns, most notably high level expression of HOX genes. The preponderance of data suggest that MLL fusion protein transformation is primarily mediated through direct upregulation of the A cluster HOX genes (18, 62, 63). Early evidence for a HOX gene role in leukemia came from studies of BXH2 mice. These mice developed spontaneous leukemia as a result of ecotropic retroviral integration. One of the most common integration sites in these leukemias was at Hoxa7 or a9, resulting in their over expression (64, 65). Subsequently, it was determined that human leukemias with MLL rearrangements consistently express high levels of HOXA7 and HOXA9 (66-69) with the rare exception of low level A cluster HOX expression in a subset of ALL cases with the t(4;11), (70). In the studies reported to date, continual expression of both Hoxa9 and Meis1 is critical for maintaining MLL fusion protein mediated immortalization (63, 71). In addition, MLL amplification with HOXA9 upregulation also appears to be an important mechanism in both MDS and AML (72-74).

When over expressed by itself, Hoxa9 is only weakly oncogenic. Only rare mice transplanted with Hoxa9 transduced bone marrow develop leukemia, and those that do arise have a latency period of 6 months or more. However, overexpression of Meis1, a homeodomain containing cofactor, greatly increases the leukemogenicity of Hoxa9 (75). More than 90% of leukemias arising in BXH2 mice with Hoxa7 or a9 overexpression have a second integration site at Meis1 resulting in its over expression (64, 65). In addition, microarray studies of human leukemias show that Meis1 is consistently expressed in leukemias that express high levels of Hoxa9, either with or without MLL rearrangement (66-69). Another potential contributor to MLL mediated transformation is the microRNA mir-196b, which is located 5’ to the Hoxa9 transcription start site (76). Mir-196b is over expressed in the majority of ALL and AML cases with MLL rearrangements. Over expression of mir-196b leads to enhanced colony forming ability and a partial block in hematopoiesis in replating assays, while mir-196b specific antagomir treatment markedly decreased replating activity (76).

In addition to deregulated HOX gene expression, MLL rearranged leukemias have been found to have a transcription profile that closely resembles embryonic (ES) as opposed to hematopoietic stem cells (HSC) (77). In particular, expression of just three ES signature genes Myb, Hmgb3, and Cbx5 is sufficient for immortalization of hematopoietic progenitors in the absence of upregulated HOX gene expression (77). The WNT signaling pathway has also been implicated in the establishment of leukemic stem cells. MLL fusion proteins are capable of transforming not only hematopoietic stem cells (HSC) but also more differentiated granulocyte-macrophage progenitors (GMP). In contrast, Hoxa9 and Meis1 transform only HSC (78). The crucial difference appears to be upregulation of prostaglandin-endoperoxide synthetase cyclooxygenase 1, and the prostaglandin receptor Ptger1 and the activation of the β-catenin pathway by the MLL fusion protein (78). Interestingly, inhibition of the WNT signaling pathway through β-catenin knockout or inhibition of β-catenin synthesis with the COX2 inhibitor indomethacin differentially suppressed growth of leukemia initiating cells and not normal HSC (78).

6. Mechanisms of transformation by MLL fusion proteins: nuclear translocation partners

More than 60 different MLL translocation partners have been identified, making it challenging to develop a single model for MLL fusion protein transformation. Nonetheless, some unifying principles have emerged. The great majority of acute leukemias with MLL translocations show highly upregulated HOXA9 and MEIS1 expression. In addition, all MLL fusion proteins retain the amino terminal part of MLL that is required for MLL association with chromatin while deleting the PHD fingers, which have an inhibitory effect on transformation (79). Furthermore, in-frame fusion of MLL to a translocation partner is required for transformation, indicating that MLL mediated transformation is not a loss of function mechanism. Finally, biochemical studies show that a number of the most common translocation partners are physically associated in complexes that are involved in transcriptional elongation(80-86).

Some of the most common MLL translocation partners including AF4, AF9, ENL, and ELL, along with the less common partner AF5q31 (also known as AFF4), have been found to associate in a large complex termed “Super Elongation Complex” by Shilatifard and colleagues that also includes the elongation factors ELL2, ELL3, the elongation factors EAF1 and EAF2 and the p-TEFb complex, which is composed of CDK9 and Cyclin T1 or T2 (82). P-TEFb is required to phosphorylate the RNA Pol II C terminal domain, which promotes transcriptional elongation. A related complex AEP (AF4/ENL/P-TEFb) reported by Cleary and colleagues lacks the ELL and EAP subunits (Figure 4) (82). Several of the subunits in these complexes have been implicated in the regulation of transcriptional initiation or elongation. AF4 and AF5q31 are members of a family of related proteins including FMR2 and LAF4, the latter of which are associated with pro-B-cell leukemia. ENL and AF9 are homologous nuclear proteins that both contain a YEATs domain implicated in histone binding and transcriptional regulation. An 84 amino acid C-terminal domain of ENL, which is highly conserved with AF9 and is retained in all MLL-ENL fusions, is sufficient for transformation when fused to MLL (87). Interestingly, a short sequence shared between the AF9 and ENL C termini has been shown to interact with other SEC/AEC components including AF4 and AF5q31 (88). ELL, for example, was originally identified as a transcription elongation factor (89) and AFF4 has been shown to directly interact with P-TEFb and to regulate its kinase activity (80).

Figure 4.

MLL complex proteins during normal and malignant hematopoiesis. MLL interacts with a variety of protein complexes in hematopoietic stem and progenitor cells to promote transcription of critical target genes like HOXA9 and MEIS1. The PAF complex associates with RNA pol II and recruits the RAD6/BRE1 E2/E3 ubiquitin ligase, which promotes mono-ubiquitination of histone H2B (Ub). H2B mono-ubiquitination is a histone mark associated with transcriptional activation. PAFc, along with Menin/LEDGF, recruit the MLL complex to target genes which delivers H3K4 (Me) methyltransferase activity and promotes gene transcription. MLL associates with the HAT MOF, which promotes further gene transcription through histone H4K16 acetylation (Ac). During hematopoietic differentiation, MLL is not recruited to target genes in part due to decreased transcription of PAFc. Insufficient recruitment of MLL leads to decreased expression of target genes. Chromosomal translocations involving MLL generate MLL fusion proteins that can recruit transcriptional activation complexes dependent on the fusion partner. MLL translocation partners including AF4, AF9, ENL, ELL and AF5q31 can form a transcriptional activation complex and a related complex AEP (AF4/ENL/P-TEFb) has also been reported. These complexes involve the recruitment of pTEFb, which is required to phosphorylate the RNA Pol II C terminal domain, which promotes transcriptional elongation. The H3K79 methyltransferase DOT1l is also recruited to some MLL fusion proteins (MLL-AF10), which can further promote transcriptional activation.

AF9 and ENL have also been shown to be components of another complex, termed Dot.Com, which includes the MLL translocation partners AF10 and AF17 along with WNT signaling components TRAPP and SKP1 (85). AF10 and AF17 are highly homologous and show a number of motifs including a PHD finger, AT hook and leucine zipper domain. Importantly, both interact with Dot1l (85, 90). Studies with Dot1l knockout mice indicate that Dot1l recruitment is important for transformation by at least a subset of MLL fusion proteins (91, 92). Dot1l is required for HOXA9 and MEIS1 upregulation, which is critical to transformation, however Dot.com is also important for WNT signaling, which may be a second pathway important for leukemogenesis (78, 85). At this point, it appears that both p-TEFb and DOT1L contribute to leukemogenesis, however, their relative contributions and to what extent non-HOX target genes are important for this remain to be determined (Figure 4).

Emerging evidence suggests that there are a number of alternative mechanisms that can result in HOX deregulation and leukemogenesis. For example in rare cases, MLL is fused to translocation partners with intrinsic histone-modifying activity including the known histone acetyltransferases (HATs) CBP or P300. In these translocation products, the CBP HAT domain, as well as an adjacent bromodomain are required for transformation (93). Transformation by another infrequent MLL translocation partner EEN has been reported to involve recruitment of the histone H4 specific arginine methyltransferase PRMT1 (94).

Increasing evidence suggests that wild type MLL plays a role in MLL fusion protein leukemogenesis. The HOX A loci in MLL rearranged leukemias have high levels of both histone H3 lysine 79 and histone H3 lysine 4 methylation (53) and activation of a conditional MLL fusion protein MLL-ENL increases MLL association and H3K4 methylation at MLL target genes (53). In keeping with this, wild type MLL has recently been shown to be required for MLL-AF9 transformation (95). This is significant as it suggest that MLL methyltransferase activity may also be a therapeutic target in leukemias with MLL rearrangements. Another intriguing and controversial finding is the possible role of reciprocal translocations, which are uncommonly expressed, in MLL rearranged leukemias. For unclear reasons the MLL-AF4 fusion protein is not leukemogenic in retroviral transduction models, however, expression of the reciprocal AF4-MLL chimera did result in development of pro B cell ALL (96) apparently via a mechanism involving recruitment of DOT1L and P-TEFb (97).

7. Mechanisms of transformation by MLL fusion proteins: dimerizing translocation partners

Despite similarities between some of the more common MLL translocations, many MLL translocation partners are cytoplasmic proteins lacking intrinsic transcriptional activity. The finding of self-association motifs in many cytoplasmic MLL translocation partners suggested that MLL dimerization is transforming. The first evidence of this was the development of leukemia in Mll–lacZ mice (58). β-Galactosidase occurs as a tetramer in solution, suggesting that MLL dimerization or oligomerization by the translocation partner is oncogenic. Subsequent experiments have shown that the dimerization of MLL contributes to transformation by a number of MLL fusion proteins including AF1p, GAS7, GEPHYRIN and AF6 (62, 98). The mechanisms by which the dimerization of truncated MLL makes it transforming are unknown. Given the multivalent nature of the MLL-menin interaction, it is possible that wild type MLL dimerizes. It is attractive to speculate that this promotes coactivator (such as Pafc or the AEP complex) recruitment. In addition, dimerized MLL fusions might alter the binding activity of wild-type MLL, thereby disrupting normal MLL regulation (53). Regardless of the mechanism, the end result of MLL fusion protein dimerization appears to be upregulation of HOXA7, HOXA9 and MEIS1 (62, 98). The dimerizing MLL fusion proteins are relatively weak oncogenes, however, so that secondary genetic events such as FLT3 internal tandem duplication appear to be particularly important for transformation by these fusion proteins (99).

8. MLL amplification and partial tandem duplication in acute leukemia

Some myelodysplastic syndrome (MDS) and AML cases are associated with increased MLL copy number, either as a result of double minute chromosomes or homogenous staining region (72-74). MLL amplification appears to be more common in older patients with complex karyotypes, often with 5q- and poor outcome (100). Unlike MLL PTD cases, MLL amplification is associated with the upregulation of at least some of the genes that are consistently expressed in leukemias with MLL rearrangements, suggesting similar mechanisms of transformation as MLL fusion proteins. These include confirmed direct targets of MLL, such as HOXA7, HOXA9 and MEIS1, as well as other potential direct or indirect targets, including PROML1, ADAM10, NKG2D and ITPA (101).

About 7.5% of AML cases with normal cytogenetics harbor internal tandem duplications of MLL spanning sequences encoding the AT hook through CXXC DNA-binding domains (102). Most of these cases occur in adults in association with trisomy II or FLT3-ITD and appear to be associated with a worse prognosis than those without MLL rearrangements (103). Studies to date indicate that the other MLL allele is silenced by DNA methylation (104). The crucial oncogenic alteration is not known but may be the duplication of DNA binding motifs in MLL, as previously we showed that molecular mimics of the MLL-PTD increase the affinity for MLL binding at a target site (62). Another possibility is that the exon-duplicated form of MLL has a conformation that interferes with its normal regulation. As is the case for dimerizing MLL fusion proteins, FLT3 activation appears to be an important cooperative event in MLL PTD-associated leukemias. About half of MLL PTD cases show FLT3-ITD or point mutation. Murine knockin models expressing the MLL-PTD show Hoxa9, but not Meis1, upregulation (105). These mice do not develop leukemia on their own but rapidly succumb to AML when crossed with Flt3-ITD transgenic mice (Caliguiri, M. unpublished data). Thus far microarray expression profiling of human leukemias with MLL PTD has not revealed this characteristic signature of HOX over expression (106) raising the possibility that PTD cases transformed through mechanisms that are different to the balanced translocations of MLL and that this probably involves genes other than Hox genes.

9. Potential therapeutic targets in MLL rearranged leukemias

Inhibiting recruitment of MLL fusion proteins to target loci

A number of different strategies to target mechanisms used by MLL fusion proteins are currently under investigation. In addition to these, targeting wild type MLL may be effective in other types of leukemia with high-level HOX expression. One potential strategy is to block MLL fusion protein or associated coactivator recruitment to target genes. Small molecular inhibitors are under development to block MLL and menin or MLL, menin and LEDGF interactions, which is required for transformation by MLL fusion proteins (44, 47, 107). Another potential therapeutic target is the CXXC zinc finger domain of MLL, a region retained in all MLL fusion proteins that is crucial for MLL binding to unmethylated CpG-rich DNA (108). The crystal structure of the CXXC domain has been determined (28, 30), which will facilitate development of small molecule inhibitors that block DNA binding. Disruption of interaction between the MLL pre CXXC and RD2 regions and the PAF complex, which are also required for MLL recruitment, is another strategy for therapeutic intervention that is currently under development (34, 48).

Inhibiting the activity or recruitment of MLL fusion protein coactivators

Recruitment of P-TEFb is a critical step in transcriptional deregulation leading to leukemia (86). For leukemias involving the common nuclear translocation partners (ENL, AF4, AF9, AF5q31) inhibition of P-TEFb CDK9 activity is one possible therapeutic strategy. This is particularly feasible given the potent and apparently specific inhibitory activity of flavinoids such as flavopiridol, which is currently in phase II trials for a variety of hematologic and solid tumors. To date, use of these compounds has been limited by both their limited efficacy and high toxicity (109, 110). We, and others, have recently shown that Dot1l is required for MLL transformation and therefore represents another attractive therapeutic target (91, 92). The finding that MLL is required for transformation by MLL-AF9 and likely other MLL fusion proteins, also raises the possibility that inhibiting MLL methyltransferase activity, either directly via the SET domain or through blocking interactions with essential core components such as WDR5, may hold promise for leukemia therapy. Remarkably, mice lacking the MLL SET domain are relatively healthy, which is an encouraging sign that this will be an efficacious drug target. Specific inhibitors of MLL methyltransferase activity might also find application in MDS and AML cases showing MLL amplification or tandem duplication as well as other leukemias with deregulation of HOX gene expression.

Inhibition of downstream or cooperating pathways

The effectors of transformation by A cluster HOX genes and MEIS1 are actively being explored and a small number of potentially important targets have been identified. One of the promising classes of compounds are GSK3β inhibitors, which by inhibiting phosphorylation of CREB1, blocks its association with the HOX/PBX/MEIS transcription complex resulting in a loss of coactivator activity (111). Work is also ongoing to test the efficacy of FLT3 inhibitors, however the results of clinical trials thus far have generally been disappointing, largely because of the acquisition of resistance in leukemic blasts (112, 113).

Glossary

Epigenetics

the study of heritable changes in gene expression caused by mechanisms other than DNA sequence

MLL

Mixed Lineage Leukemia, histone methyltransferase

ALL

Acute Lymphoid Leukemia

AML

Acute Myeloid Leukemia

MDS

Myelodysplastic Syndrome

AF4

ALL1-fused gene from chromosome 4 (MLL translocation partner)

AF9

ALL1-fused gene from chromosome 9 (MLL translocation partner)

ENL

Eleven Nineteen Leukemia (MLL translocation partner)

AF10

ALL1-fused gene from chromosome 10 (MLL translocation partner)

AF6

ALL1-fused gene from chromosome 6 (MLL translocation partner)

ELL

RNA polymerase elongation factor (MLL translocation partner)

AF1p

Epidermal growth factor receptor pathway substrate 15 (MLL translocation partner)

AF17

ALL1-fused gene from chromosome 17 (MLL translocation partner)

SEPT6

Septin 6 (MLL translocation partner)

AT-Hooks

Evolutionarily conserved domain of MLL that binds to AT rich DNA

CBP

cAMP response-element binding protein (CREB) binding protein. A transcriptional coactivator with histone acetyltransferase activity (MLL translocation partner)

Cyp33

Cyclophilin 33, an RNA binding protein with proline isomerase activity

HDAC

Histone Deactetylase. Enzymes that remove acetyl groups from histone and associated with transcriptional repression.

PHD finger

Plant Homeodomain finger. Structural domain that can bind methylated histones

PcG

Polycomb group proteins. Involved in transcriptional repression

SNL

Subnuclear localization domain

ASH2l

absent, small, or homeotic-like (Drosophila), interacts with MLL

WDR5

WD repeat-containing protein 5, interacts with MLL

RbBP5

Retinoblastoma-binding protein 5, interacts with MLL

MOF

(MYST1) ortholog of Drosophila males absent on the first, histone acetyltransferase that interacts with MLL

HCF

Host Cell Factor, interacts with MLL

Menin

(MEN1) Tumor suppressor associated with multiple endocrine neoplasia type 1

LEDGF

(Psip1) Lens epithelium derived growth factor, chromatin associated transcriptional co-activator, interacts with MLL

PAFc

Polymerase Associated Factor complex, interacts with MLL

DOT1l

Disruptor of telomeric silencing (Yeast) Histone H3 methyltransferase

p-TEFb

Positive Transcription Elongation Factor b, composed of CDK9 and one of either Cyclin T1, T2 or K. Kinase activity phosphorylates serine 2 of RNA pol II CTD