The MLL partial tandem duplication: evidence for recessive gain-of-function in acute myeloid leukemia identifies a novel patient subgroup for molecular-targeted therapy

Cytogenetic abnormalities involving chromosome band 11q23, such as translocations, deletions, and duplications, are seen in approximately 15% of patients with acute myeloid leukemia (AML) and most often result in gene fusions between the 5′-end of the MLL (ALL-1) gene and the 3′-end of a partner gene.¹  One type of MLL rearrangement not detectable by classic cytogenetics is the partial tandem duplication of MLL (MLL PTD). This rearrangement most commonly results from a duplication of a genomic region encompassing either MLL exons 5 through 11 or MLL exons 5 through 12 that is inserted into intron 4 of a full-length MLL gene, thus fusing introns 11 or 12 with intron 4 (exon designations used throughout are consistent with GenBank NT_033899.6). At the level of transcription this results in a unique in-frame fusion of exons 11 or 12 upstream of exon 5.^2,3  The presence of an MLL PTD at the DNA and RNA level has been demonstrated most often in adult de novo AML with normal cytogenetics or trisomy 11 (+11), but it has also been observed in childhood leukemias, adult acute lymphoblastic leukemia (ALL), secondary leukemia, and a solid tumor cell line.^4-6  In adult de novo AML with a normal karyotype, the presence of the MLL PTD has been associated with a worse prognosis (ie, shorter duration of remission) when compared with normal karyotype AML without the MLL PTD.^7-9 

The MLL PTD self-fusion has a duplicated N-terminal region that contains the AT hook DNA-binding motifs, a domain that preferentially binds an unmethylated cytosine in cytidine phosphate guanosine (CpG) dinucleotides on DNA, and a transcriptional repression domain.^3,10  In the absence of a fusion gene partner that would replace the 3′-end of the transcript, the mechanism by which this PTD functions in leukemogenesis, drug resistance, and leukemia relapse is currently unknown. We previously used Southern analysis to demonstrate that the MLL PTD defect is present on only one chromosome 11 and the other has a wild-type (WT) MLL allele in patients with AML with a normal karyotype. Likewise, in AML with +11 and the MLL PTD, 2 chromosomes 11 contain MLL WT alleles, while the third copy contains the MLL PTD.² 

To examine the relationship between the MLL PTD and the MLL WT and to better understand the importance of the MLL PTD in leukemogenesis, we quantified each of these 2 gene products in primary AML blasts containing this molecular defect. Our results show the MLL WT transcript is not expressed in primary AML blasts that harbor the MLL PTD, in contrast to AML with either MLL WT genes only or the t(9;11)(p22;q23). Induction of MLL WT in response to a DNA methyltransferase (DNMT) inhibitor and/or a histone deacetylase (HDAC) inhibitor in these cases was selectively associated with enhanced sensitivity to cell death. As the absence of MLL WT protein is predicted to contribute to the leukemic phenotype, these data appear to identify a new molecular target for DNMT or HDAC inhibitors or both in patients with AML with the MLL^PTD/WT genotype.

Diagnostic and, where indicated, subsequent relapse blood and bone marrow samples were obtained from patients with AML. All patients gave written informed consent on protocols approved by separate institutional review boards (IRBs) for the storage and use of procured cells in research studies (Cancer and Leukemia Group B [CALGB] 9665 and 9769 and Ohio State University [OSU] 1997C0194). Percentage of blasts prior to enrichment was determined from Wright-Giemsa–stained cells. Mononuclear cells were enriched through a Ficoll Hypaque gradient and then cryopreserved. Karyotyping was performed and centrally reviewed by the CALGB (CALGB 8461).¹¹  Peripheral blood and CD34⁺ hematopoietic progenitor cells from normal bone marrow cells were obtained with written IRB-approved consent from healthy donors. CD34⁺ hematopoietic progenitor cells (HPCs) were isolated to 97% or greater purity using immunomagnetic beads (Miltenyi Biotec GmBH, Cologne, Germany) followed by fluorescence-activated cell sorting as described elsewhere.¹² 

Cells were quickly thawed and viability was greater than 90%. Conventional nested reverse transcription–polymerase chain reaction (RT-PCR) was performed as previously described.¹³  The cloned PCR products were sequenced (The Ohio State University Comprehensive Cancer Center's Genotyping and Sequencing Shared Resource). Only samples in which one MLL PTD-derived transcript was observed by nested RT-PCR were used in the study; those with splice variants detected were excluded. All patient samples with the MLL^PTD/WT genotype had MLL gene rearrangement verified by Southern blotting as described previously using probes B859 and SAS1.³ 

Primer pairs and dual-labeled probe sets were designed to amplify sites that are unique to the MLL PTD or common to both the MLL PTD and the MLL WT transcripts. Primer and probe sets were designed to amplify the “unique amplicons” exon 11 to exon 5 or exon 12 to exon 5 fusions specific for the 2 most common forms of the MLL PTD, and to amplify the “common amplicons” exon 11 to exon 12, exon 13 to exon 14, and exon 26 to exon 27 junctions that can be found in both the MLL PTD and the MLL WT transcripts (Figure 1). QRT-PCR amplification of 18S rRNA or ABL was performed to normalize for differences in the amount of total nucleic acid added to the reaction. Standard curves were constructed to allow for measurement of target amplicon copy numbers. QRT-PCR data were collected using the ABI Prism 7700 Sequence Detection System (PE Applied Biosystems, Foster City, CA). All assays were optimized and corrected for any differences in PCR efficiencies to accurately reflect allele-specific expression. The lower limit of detection was determined to be 10 amplicon target copies in 125 ng reverse transcribed total RNA. Primers are available upon request.

Immunoblotting analysis for detection of the p300-kDa MLL WT and predicted approximate p420-kDa MLL PTD N-terminal fragments was carried out as previously described.¹⁴  Briefly, 40 μg nuclear extracts was size fractionated in a 4.9% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). After transfer, membranes were probed with anti-MLL 170 antibody, an affinity-purified anti-MLL antibody directed against the N-terminal p300 MLL WT posttranslational cleavage product (a generous gift from Dr Tatsuya Nakamura, Ohio State University).¹⁴  Proteins were visualized using enhanced chemiluminescence Plus (Amersham-Pharmacia, Piscataway, NJ).

Cryopreserved blasts from patients with AML were thawed, washed in fetal calf serum, and cultured at a concentration of 5 × 10⁵ cells/mL in improved minimal essential medium (IMEM) supplemented with 10% fetal calf serum (FCS), 10 ng/mL PIXY321 (recombinant interleukin-3 and granulocyte-macrophage colony-stimulating factor [rIL-3/GM-CSF]), and 10 ng/mL stem cell factor (Amgen, Thousand Oaks, CA). After overnight incubation, some of the cells were cultured with media alone or 20.5 μM of the DNMT inhibitor, 5′-aza-2′-deoxycytidine (5′-Aza-CdR or decitabine; Sigma, St Louis, MO) for 48 hours.¹⁵  Cultures were replenished with fresh media containing either 0 or 3 nM of the HDAC inhibitor, depsipeptide (FR901228; Developmental Therapeutics Program, Division of Cancer Treatment, National Cancer Institute, Bethesda, MD), and incubated for an additional time as indicated.

The MLL 5′-CpG islands were identified using the algorithm described in http://www.ebi.ac.uk/emboss/cpgplot/ and MLL genomic sequence (NCBI GenBank Accession No. NT033899.6). Methylation status was assessed by bisulfite PCR sequencing (BS-PCR) of genomic DNA as previously described.¹⁶  PCRs were first optimized to minimize the potential for bias toward amplification of nonmethylated sequences. Single PCR products were then purified from the agarose gel, cloned into the pCR2.1 cloning vector (Invitrogen, Carlsbad, CA), and sequenced. A minimum of 10 clones per PCR were evaluated.

ChIP assays were performed according to the manufacturer's suggested protocols using antibodies specific to histone H4 acetylated at K5, K8, K12, and K16; histone H3 acetylated at K9 and K14; and histone H3 methylated at K9 (Upstate Biotechnology, Waltham, MA). An MLL-specific primer pair was designed to amplify a region upstream of the transcriptional initiation site in MLL (nucleotides -168 to -2). For normalization, ChIP analysis of the housekeeping gene, GAPDH, was performed using GAPDH promoter-specific primers previously described.¹⁷  PCR conditions were optimized such that products were detected during the exponential phase of amplification. Relative quantification was carried out using SybrGreen dye and real-time PCR. The comparative real-time PCR (2^-ΔΔCT) method was used, normalizing first to input DNA followed by depsipeptide-treated levels relative to control levels.

Three 19-mer oligodeoxynucleotides (ODNs) were designed, and all had a phosphorothioate backbone. The antisense ODN (aODN) spans the fusion of MLL exon 12 with exon 5 present in the MLL^PTD/WT primary AML sample evaluated. The control sense ODN (sODN) is complementary to the aODN, and the control scrambled ODN (scrODN) has the same nucleotide content as the aODN but in a scrambled sequence. Primers are available upon request. Leukemic blasts were plated in 24-well plates at a concentration of 5 × 10⁵ cells/well in Iscoves modified Dulbecco medium (IMDM) supplemented with 10% fetal bovine serum (FBS), 10 ng/mL stem cell factor, 10 ng/mL PIXY321 and incubated at 37°C, 5% CO₂ for 2 hours before additional incubation with no ODN (media control), the sense, scrambled, or the antisense ODNs (10 μg/mL). To ensure constant ODN delivery to those cells that are transcribing new MLL mRNA, 24 and 48 hours after plating, 50% of the original ODN quantity was re-added to the appropriate wells. Total RNA was extracted from 2.5 × 10⁵ cells after 48 hours of incubation with ODNs. QRT-PCR was performed to quantify MLL PTD and MLL WT transcripts.

After 72 hours, 1 × 10⁵ cells were harvested from each well and plated in duplicate in methylcellulose cultures (MethoCult; Stem Cell Technologies, Vancouver, BC, Canada). Plates were incubated at 37°C, 5% CO₂ until colonies (> 50 cells) were observed in media control plates. Colonies were then counted on each plate, and the mean ± SD of the 2 plates from each condition were recorded and plotted. The number of colonies arising in the plates that contained no ODNs represented maximal (ie, 100%) growth for that individual. The numbers of colonies arising in the ODN-treated plates were expressed as a percentage of this number.

To evaluate the impact of 5′-Aza-CdR, depsipeptide, or the sequential combination treatment on MLL WT transcript levels relative to the MLL PTD transcript levels, the Wilcoxon signed rank test was used. The Mann-Whitney U test was used to compare whether the cell death levels induced by the combination treatment and relative to the control samples were significantly different between MLL^PTD/WT and MLL^WT/WT AML samples. All tests of statistical significance were 2-sided at an α-level of .05.

We developed a modified allele-specific transcript discrimination assay based on the Taqman real-time PCR assay and measured MLL allele–derived transcripts in primary AML blasts from 19 patients with the MLL PTD (17 with normal cytogenetics and 2 with trisomy 11 as a sole abnormality). To determine transcript copy number, the common amplicon (shared between both WT and PTD alleles) copy number was subtracted from the MLL PTD unique fusion amplicon copy number. Both the MLL PTD exons 5 to 11 and MLL WT transcripts have the same stability (S.P.W., unpublished data, July 2004), so if both alleles are transcribed at equivalent rates, the common-to-unique amplicon copy number ratio would be 2:1 in an AML sample with the MLL^PTD/WT genotype and a PTD of exons 5 through 11. If the ratio observed is 1:1, this implies the MLL PTD allele is expressed and the MLL WT allele is silent or missing (Figure 1; Table 1). Although the presence of MLL WT allele (or alleles as in the cases with +11) was confirmed at the genomic level by Southern blot, the MLL WT transcript was absent or minimally detected by our QRT-PCR assay in 18 of 19 MLL PTD-positive AML samples (Table 1). An MLL PTD–positive patient with trisomy 11 that also expressed MLL WT at diagnosis lost expression of the WT transcript at relapse (Table 1, unique patient number [UPN] 023d [diagnosis] and 023r [relapse]). None of the 12 primary AML patient samples with the MLL^WT/WT genotype and normal karyotypes had any measurable MLL PTD transcript, yet all expressed high levels of the MLL WT transcript using our QRT-PCR assay (Table 1).

Long-range RT-PCR spanning from the MLL PTD exon 11 to 5 fusion site to exon 12 followed by sequencing analysis supported that both the common and unique amplicons were derived from the same transcript, the MLL PTD, in primary AML samples harboring the MLL PTD of exons 5 through 11 (data not shown). Consistent with this, the predicted N-terminal MLL posttranslational cleavage protein product¹⁸  of the MLL PTD (∼ p420 kDa) was detected in the MLL PTD–positive Mgc80-3 cell line⁴  and in a MLL^PTD/WT genotype primary AML sample (UPN 300) (Figure 2, lanes 1-2). Unlike the Mgc80-3 cells, and consistent with the QRT-PCR data, no evidence of the p300 N-terminal protein fragment arising from cleavage of the full-length MLL WT protein was seen in this primary AML sample (Figure 2, lane 2). As expected, the p420-kDa fragment is absent, while the p300-kDa protein is present in a primary AML sample (UPN 003) and in the K562 cell line, both harboring the MLL^WT/WT genotype and expressing only MLL WT transcript (Figure 2, lanes 3-4).

To determine whether absence of the MLL WT transcript is a common occurrence in MLL-associated leukemia in which MLL PTD is not present, we performed QRT-PCR using AML blasts from 4 patients each harboring the t(9;11)(p22;q23) that fuses MLL with AF9. The MLL-AF9 chimeric transcript specific for each of the 4 AML cases, ie, amplification of MLL exon 11 to AF9 exon 5 fusion, and the MLL WT transcript, ie, the exon 11 to exon 12 amplicon that is absent in the chimeric mRNAs, were present in all 4 samples. Indeed, the MLL WT/MLL-AF9 transcript ratio varied from nearly equivalent levels (ratio = 0.87) to a ratio of 11.8 (Table 2). These data suggest that silencing of the MLL WT allelic expression is unique to MLL^PTD/WT AML.

We speculated that the inactive MLL WT allele could be silenced by epigenetic modifications and, thus, pharmacologically induced in AML cases with the MLL^PTD/WT genotype using either DNMT or HDAC inhibitors. Incubation of primary MLL^PTD/WT AML blasts that lacked basal MLL WT expression with either the DNMT inhibitor, 5′-Aza-CdR (0.5 μM), or the HDAC inhibitor, depsipeptide (3 nM), modestly induced MLL WT transcripts as evidenced by the increase in the MLL WT/MLL PTD ratio (Figure 3). The combination of 5′-Aza-CdR and depsipeptide resulted in the highest MLL WT/MLL PTD ratio in 6 of 7 primary AML blasts with the MLL^PTD/WT genotype as compared with either agent alone (Figure 3). While the agents singly or in combination induced MLL WT relative to the media-only control group, pairwise comparisons indicated the MLL WT/PTD ratio measured in the combination regimen group was significantly higher than in the 5′-Aza-CdR group but only borderline significant relative to the depsipeptide treatment group (Wilcoxon signed rank test, P = .031 and P = .063, respectively). In contrast, there was no significant change in the level of MLL WT transcript relative to the levels of 18S and ABL transcript in 5 of 5 primary AML samples with high blast percentages and the MLL^WT/WT genotype that were treated in vitro with 5′-Aza-CdR, depsipeptide, or both drugs in combination (data not shown).

We next investigated whether direct DNA hypermethylation of the MLL WT promoter, chromatin remodeling, or both could be contributing to MLL WT gene silencing. First, we examined the methylation status of the 2 CpG islands located upstream and encompassing the MLL gene's transcriptional initiation site (279 base pair [bp] in length [-1156 to -878] and 1.08 kb in length [-527 to +548]). Bisulfite-PCR (Figure 4) performed on 2 primary MLL^PTD/WT AML samples with normal cytogenetics and a high percentage of blasts (UPNs 300 and 316), indicated a sporadic, low level of methylation (1%-34%) in 48-hour cultures similar to the methylation levels detected in a primary MLL^WT/WT AML sample, in normal CD34⁺ bone marrow cells, and in peripheral blood mononuclear cells from disease-free donors. As expected, incubation of the 2 MLL^PTD/WT AML samples with 5′-Aza-CdR reduced methylation at several sites (data not shown). The data suggested that preferential methylation of the MLL WT allele was not responsible for the silencing of the MLL WT allele in MLL^PTD/WT AML samples.

Chromatin remodeling by modified histones is a second epigenetic mechanism that could silence the MLL WT gene. Transcriptional repressive histone modifications, eg, deacetylated histones H3 and H4, are known to silence gene expression, and acetylated histones present along a promoter region have been associated with activation of gene expression.¹⁹  We, therefore, examined the status of acetylated histones bound near the MLL transcriptional initiation site in primary AML blasts incubated with or without depsipeptide. The baseline levels of bound acetylated histone H3 and acetylated histone H4 normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels were different between the MLL^PTD/WT (UPN 300) and the MLL^WT/WT (UPN 003) primary AML blasts (Figure 5). Depsipeptide induced an early increase of acetylated histones H3 and H4, approximately 300% and 1200%, respectively, above the basal levels in the media-only controls in the MLL^PTD/WT sample (Figure 5A). Consistent with the change in histone H3 and H4 status occurring on the MLL WT allele in the MLL^PTD/WT sample, we observed induced MLL WT expression but not MLL PTD levels when the leukemic blasts were incubated longer with drug (data not shown). In contrast, levels of acetylated histones H3 and H4 were increased after depsipeptide treatment by less than 20% relative to the media-only control in the depsipeptide-treated primary AML blasts with the MLL^WT/WT genotype (Figure 5B), and no change in MLL WT transcript level was observed after depsipeptide treatment (data not shown). The ChIP-PCR gel results for each primary AML sample were confirmed by relative quantification with the SybrGreen real-time PCR assay (Applied Biosystems, Foster City, CA) after ChIP (Figure 5C).

To determine whether the induction of MLL WT transcript leads to functional protein production and predictable downstream changes, we assayed for the expression of the MLL downstream target, HOXA9, in MLL^PTD/WT AML cells treated with a combination of 5′-Aza-CdR and depsipeptide. Indeed, when we observed a greater than 4-fold increase in MLL WT, HOXA9 mRNA levels increased 2.5-fold relative to media-only control cells. In contrast, when the combination failed to increase MLL WT in an MLL^PTD/WT AML sample, no increase in HOXA9 was seen (data not shown).

Inhibition of DNMT and HDAC activities by 5′-Aza-CdR and depsipeptide, respectively, has been associated with tumor cell death.²⁰  Therefore, we assessed whether treatment with these compounds in our experiments affected cell viability as determined by trypan blue exclusion. Reduction in viable cells was observed in 7 of 8 primary AML bone marrow samples with the MLL^PTD/WT genotype after in vitro treatment with the combination of 5′-Aza-CdR and depsipeptide (Figure 6). One primary MLL^PTD/WT AML sample (UPN 146) that was refractory to cell death also did not exhibit induction of MLL WT mRNA levels after treatment with the combination. The same drug combination was not effective in 6 of 7 MLL^WT/WT AML samples (Figure 6). This difference in cell death between the 2 genotypes was significant (P = .021).

To further evaluate whether the MLL PTD and MLL WT have opposing functions in the context of leukemic blast survival/proliferation, we down-modulated MLL PTD expression in primary AML blasts using aODNs directed against the unique MLL PTD transcript fusion site. The 19-mer MLL PTD aODN decreased levels of this fusion transcript. Notably, treatment of AML blasts with MLL PTD aODN was associated with a ratio of MLL WT to MLL PTD transcript of 1.85. As the MLL WT transcript was not detected in the nontreated cells, the measured MLL WT/MLL PTD ratio supports the conclusion of an absolute increase in MLL WT rather than an apparent alteration in the ratio of MLL WT/MLL PTD. Neither the sense (sODN) nor scrambled (scrODN) ODN decreased the levels of the MLL PTD transcript or induced MLL WT expression (Figure 7A).

To examine the functional effectiveness of aODNs directed toward the MLL PTD, AML-CFU assays were performed using the same experimental conditions as described above. The aODNs directed against the unique MLL PTD exon 12 to exon 5 fusion present in the primary MLL^PTD/WT AML sample inhibited AML-CFU blast colony formation (maximum, 95% inhibition), while the percentage of inhibition in the presence of control ODNs was negligible (Figure 7B). Wright-Giemsa staining indicated no significant changes in morphology that would be indicative of maturation (data not shown). Similar QRT-PCR and CFU results were obtained with aODNs directed toward the fusion transcript present in primary MLL^PTD/WT AML blasts harboring the exon 5 to exon 11 PTD (data not shown).

The MLL PTD is uniquely distinguished from other MLL gene rearrangements that result in chimeric gene fusions, in that the MLL PTD retains all of the encoded domains of the MLL WT. In this study we sought to gain insight into the role of the MLL PTD in AML by assessing the relative expression levels of the MLL PTD compared with MLL WT at the mRNA level in primary AML blasts. In contrast to primary AML blasts expressing both an MLL-AF9 fusion transcript and MLL WT transcript, we observed the MLL WT transcript derived from the non-rearranged MLL allele is absent in the vast majority of MLL^PTD/WT primary AML blast samples evaluated. Examples exist, such as p53 and retinoblastoma (Rb) genes,^21,22  and the more recently described, FLT3^ITD/- genotype,²³  whereby a common mechanism for gain-of-function is a functional mutation on one allele concurrent with a deletion or suppression of the other normal allele. The virtual absence of MLL WT expression in nontreated MLL PTD⁺ AML blasts and the finding of a positive association between MLL WT induction and enhanced sensitivity to cell death is consistent with recessive gain-of-function by which the MLL PTD contributes to the leukemic phenotype. This role for the MLL PTD was further supported by reduced AML-CFUs after targeted down-regulation of the MLL PTD. Of potential interest is the induction of MLL WT mRNA levels observed with antisense-mediated decrease in MLL PTD expression, which indicates MLL PTD protein may contribute to silencing of the MLL WT allele via a possible autoregulatory mechanism.^24-26 

Additionally, the coincident suppression of the MLL WT allele with expression of the MLL PTD supports the hypothesis that loss of MLL WT function via monoallelic repression contributes to the leukemic phenotype by the remaining mutant allele. Notably, in one diagnostic AML sample with trisomy 11 and the MLL PTD, both the PTD and WT transcripts were present at nearly equivalent levels; however, only the PTD transcript was detected in the relapse sample from this AML case. This could have resulted from deletion of the expressing MLL WT allele or acquired silencing of 2 MLL WT alleles. These data, however, are supportive of a possible tumor suppressor role for the MLL WT protein and suggest that its decrease may lead to a more aggressive phenotype in the context of the MLL PTD AML subtype. MLL5 protein, another MLL gene family member located on chromosome 7q22, a region that is commonly deleted in myeloid malignancies, has been recently demonstrated to induce growth suppression and cell-cycle arrest in in vitro assays.²⁷ 

Mll^wt/- mice have a defect in their hematopoietic compartment, consistent with a gene dosage effect and suggesting that 2 WT MLL genes are required for normal hematopoiesis.²⁸  In cases whereby a PTD of MLL has occurred on one allele, loss of functional MLL WT from the other chromosome(s) 11 may further contribute to abnormal hematopoietic cell function. A gene expression profiling study comparing homozygous murine Mll knock-out and normal Mll fibroblasts further supports the idea that loss of MLL WT can contribute to abnormal cellular processes.²⁹  The researchers suggested that perhaps fusion proteins resulting from MLL translocations might compete with MLL WT protein for binding in a multiprotein complex, which in turn could alter or prohibit normal function and target gene expression. As the MLL PTD is a self-fusion containing the same and duplicated functional domains as the MLL WT, one could envision that altering normal function, eg, enhanced binding to a target gene promoter or binding to gene promoters that are not normally recognized by MLL WT, could be occurring within MLL^PTD/WT AML blasts.

The 5′-regions for the MLL PTD and MLL WT genes are identical; therefore, why the differential expression? One reasonable explanation is there could be differential protein binding at the identical nucleotide sequence with the promoter regions of each MLL allele, analogous to that demonstrated for the BCR and BCR-ABL alleles in chronic myelogenous leukemia.³⁰  This differential binding could also arise by epigenetic mechanisms regulating the 2 alleles differently. We observed that the MLL WT gene could be activated by in vitro treatment of primary MLL^PTD/WT AML blasts with DNMT and HDAC inhibitors. Epigenetic regulation mediated by DNA hypermethylation and histone deacetylation are contributors to gene silencing.^31-33  The induction of MLL WT expression observed in the majority of the MLL^PTD/WT AML blasts treated with the DNMT and HDAC inhibitor combination is consistent with direct or indirect involvement of these enzymatic activities in the suppression of the MLL WT allele. Contrary to other genes directly silenced via hypermethylation of their 5′-CpG island, the MLL 5′-CpG island is not hypermethylated in MLL^PTD/WT AML samples. Yet, the MLL WT allele could be activated subsequent to a 48-hour treatment with the DNMT inhibitor. At least 3 possibilities could explain the effects of DNMT inhibition on MLL WT. First, it is possible that other sites distant from the CpG island within the proximal promoter of MLL could be directly regulated by methylation/demethylation as shown for other genes.^34-36  In this regard, we have demonstrated a trend toward increased overall methylation on chromosome 11, relative to other chromosomes, in adult de novo AML.³⁷  Second, the induction of MLL WT mRNA observed after treatment with 5′-Aza-CdR alone and the enhancement observed in combination with depsipeptide could be a result of regulation of the MLL WT allele by trans-acting factor(s) that are themselves directly regulated by DNA methylation, a similar mechanism as suggested by others.^38,39  Finally, consistent with our observation that 5′-Aza-CdR reduced methylation of several sites within the MLL transcriptional initiation site of the MLL^PTD/WT AML samples evaluated, is the possibility that a few select sites, when methylated, inhibit expression of MLL WT, as indicated for other genes.^40-43 

Alternatively, silencing of one allele could result from differential histone status within the coding region of the MLL WT versus MLL PTD gene. Evidence for regulation of gene expression by modified histones residing in the coding region has been demonstrated.⁴⁴  One could reasonably hypothesize that, by virtue of the presence of the duplicated MLL region, binding of chromatin factors to the promoters could be different between the MLL PTD and MLL WT alleles and thus have an impact on chromatin structure that, in turn, could differentially affect the expression of the MLL PTD and WT alleles in the AML blasts. The absence of change in MLL WT gene expression in MLL^WT/WT AML cases treated with the HDAC inhibitor is consistent with the absence of change observed in levels of bound acetylated histones H3 and H4 in the 5′-region compared with control cells. While the relevance of this association is strengthened by the contrasting results in the MLL^PTD/WT AML cases, the ChIP experimental design used herein represents only what is occurring at a single time point within a single region of the MLL promoter. A greater number of patient samples are needed to gain further insight into how chromatin factors, such as histones and their modifications, contribute to regulation of the MLL alleles in AML with the MLL^PTD/WT genotype as well as in other subgroups of AML.

Finally, evidence supports an important epigenetic role for the C-terminal SET (suppressor of variegation, enhancer of zeste, trithorax) domain of MLL in chromatin remodeling via histone lysine 4 methyltransferase activity as well as recruitment of proteins involved in chromatin remodeling.^14,45,46  In contrast to the more common balanced translocations involving MLL, the MLL PTD retains the SET domain. Indeed, in recent expression profiling studies, clustering analyses demonstrate that patient AML samples with the MLL PTD have different overall expression patterns when compared with those AML samples harboring MLL chimeric fusions.^47,48  The differences in expression patterns could be due to, at least in part, the underlying differences in leukemogenic function between blasts expressing MLL chimeric fusions concurrent with MLL WT and blasts expressing MLL PTD but not MLL WT, as our data now suggest. To further understand the role of the MLL PTD gene rearrangement in AML, it will be of interest to further elucidate the effects of simultaneous overexpression of MLL PTD and MLL WT proteins in in vitro and in vivo systems on, for example, the expression of HOX genes and other downstream targets and the potential role for MLL WT protein in leukemic cell death and growth inhibition. Our results appear to suggest that one can dissociate cell death from HOXA9 up-regulation induced by MLL WT expression, at least in the MLL PTD AML subgroup. Ultimately, enhanced understanding of the relationship between MLL WT and its numerous gene rearrangements in leukemogenesis may aid in the development of important therapies for patients harboring MLL aberrations. Our results advocate the combination of a DNMT and HDAC inhibitor as one potential therapeutic strategy that might be beneficial in MLL^PTD/WT AML patients, with or without standard chemotherapy regimens. This strategy may also have special clinical utility for those patients with AML with MLL^PTD/WT/WT with trisomy 11, as prognosis for this subgroup is poor.^49,50 

We thank Dr Danilo Perrotti and Dr Tatsuya Nakamura for their valuable insights and technical expertise.