A role for the MLL fusion partner ENL in transcriptional elongation and chromatin modification

Hematopoietic development is controlled by an intricate network of finely tuned transcriptional programs. Consequently, a perturbation of the transcription factors involved can block differentiation. This developmental roadblock cooperates with mutations in pathways that signal growth and/or survival to cause acute leukemias.¹  A prime example for such a mechanism is mixed lineage leukemia. In this disease, the gene for the histone methyltransferase MLL participates in chromosomal translocations that eventually create maturation-blocking and therefore leukemogenic MLL fusion proteins.^2,3  These protein chimeras consist of N-terminal portions of MLL joined to a variety of mostly unrelated fusion partners that replace the original MLL C-terminus, including the methyltransferase domain (http://atlasgeneticsoncology.org/Genes/MLL.html). MLL fusion proteins are aberrant transcription factors that ectopically activate genes important for hematopoietic development like the abdominal-type Hox genes Hoxa7 and Hoxa9 and their dimerization partner Meis1.^4,,–7 

Despite intensive study, little is known about the biological function of MLL fusion partners in normal cells, and it is mostly unclear how these proteins activate the oncogenic potential of MLL. In the rare cases where MLL is joined to a cytoplasmatic protein, domains introduced by the partner force a dimerization of the fusion that is crucial for oncogenic activity.^8,9  The overwhelming majority of leukemias with MLL rearrangement, however, involves nuclear proteins as translocation partners. The data available seem to indicate a role of some nuclear MLL partners in transcriptional control and histone modification. Support for this speculation comes from the detection of direct protein-protein interactions of the homologous MLL fusion partners AF4 and AF5q31 that both bind to ENL and the closely related AF9.^10,11  ENL in turn interacts with histone H3.¹¹  In addition, another MLL fusion partner, AF10, recruits the histone H3 lysine 79–specific methyltransferase DOT1L that introduces a dimethyl mark during transcriptional elongation.¹²  The same modification is also a hallmark of genes activated by MLL-ENL.¹³  Finally, the proteins CBX8 (chromobox 8) and BCoR (BCL6 corepressor) that are involved in chromatin-dependent gene repression have also been found to associate with ENL and AF9.^14,–16 

In order to learn more about the biological function of a classical MLL fusion partner, we identified proteins associated with the “Eleven Nineteen Leukemia” protein (ENL) originally discovered as an MLL fusion partner in the recurrent translocation t(11;19). Here, we describe the purification and analysis of ENL-associated proteins (EAPs) by tandem immunoprecipitation of ENL. This protein assembly contains several other MLL fusion partners, positive transcription elongation factor b (pTEFb), DOT1L, and polycomb group proteins. The composition of EAP suggests that ENL works in a new unit of transcriptional regulation that coordinates transcriptional elongation with chromatin modification.

The cDNA for flagENL (GenBank accession no. NM_005934¹⁷ ), MLL-ENL, and mutants thereof were inserted into the retroviral vectorpMSCVneo. For production of anti-ENL small inhibitory (si) RNA, an oligonucleotide corresponding to ENL positions 489 to 509 (aattccactctctgccttctc) was introduced into the “self-inactivating” retroviral vector pSIRENretroQ (Clontech, Palo Alto, CA). This construct provides a source of small hairpin (sh) RNA for intracellular production of specific siRNA. Derivatives of ENL and mouse Dot1l (a generous gift of B. C. Kone, Houston, TX) for 2-hybrid experiments were cloned either into pGBKT7 as bait construct or pGADT7 for target clones. Fusions with enhanced green fluorescent protein (EGFP) or red fluorescent protein (RFP) were done in pEGFP and pRedExpress (all vectors from Clontech). The luciferase reporter plasmids have been described.⁴  Primer sequences for qPCR of HOXA9 sequences are available on request.

The 293fENL producer cell line was generated from HEK293 cells by amphotropic transduction with pMSCVneo-fENL viruses.

Anti-flag (M2) agarose and anti-flag antibody were from SIGMA (Taufkirchen, Germany). Monoclonal antibodies against ENL (clones 3.1, 4.1, and 33.1) were produced in-house. Other antibodies were from AbCam (Cambridge, United Kingdom), Santa Cruz Biotech (Santa Cruz, CA), or AbGent (San Diego, CA). Quantification of Western blots was done densitometrically with QuantityOne software (BioRad, München, Germany).

ENL affinity resin was produced by coupling monoclonal anti-ENL antibodies 3.1, 4.1, and 33.1 to a solid support with SEIZE primary immunoprecipitation reagents (Pierce, Rockford, IL). Extracts from approximately 5 × 10⁸ cells (293fENL or HEK293) were prepared by lysis with bufferP (300 mM NaCl, 20 mM HEPES [pH 7.5], 0.5 mM EDTA, 0.1% Triton X-100, 0.5 mM sodium vanadate, 2 mM NaF, 2 mM DTT, 0.2 mM PMSF, 20 μg/mL leupeptin, 0.4μg/mL aprotinin, and 40 μg/mL pepstatin A). After preclearing with protein A/G agarose (Santa Cruz Biotech) the extracts were incubated overnight with flagM2 agarose. The anti-flag beads were washed in bufferP supplemented with 0.25% bovine serum albumin. Immobilized proteins were eluted with 150 μg/mL flag-peptide (SIGMA) followed by a second precipitation with anti-ENL resin, a wash step in bufferP, and elution by low pH (100 mM glycine [pH 2.7]) and neutralization. Proteins were analyzed by SDS-PAGE and Coomassie and silver staining (ProteoSilver plus; SIGMA). Nucleases were added at 5 U/mL (benzonase) and 0.5 μg/mL (RNAseA).

Gel filtration was done with primary extracts in bufferP (without Triton X-100) on a calibrated Sephacryl S300 column (GE Healthcare, Munich, Germany).

Silver-stained bands were excised from polyacrylamide gels and proteins in each band were digested with trypsin, and the resulting peptide mixtures were analyzed by reverse-phase liquid chromatography–tandem mass spectrometry (LC-MS/MS) using a Thermo Electron (Waltham, MA) linear ion trap (LTQ) mass spectrometer. Peptides were identified by searching acquired MS/MS spectra using Mascot¹⁸  against the Human International Protein Index (IPI) protein sequence database version 3.16 with the following search parameters: monoisotopic precursor ion mass with 3 Da mass tolerance, unconstrained search, and allowing for methionine oxidation as a variable modification. Peptide assignments to spectra were statistically validated using PeptideProphet,¹⁹  which computes for each peptide assignment in the dataset a probability of being correct given its database search score, the number of tryptic termini and missed cleavages, and other properties. Peptide identifications were assembled into proteins using the protein inference tool ProteinProphet,²⁰  and the probabilities were recomputed at the protein level. All proteins with a probability of being correct greater than 0.9 were retained, which corresponds to the false discovery error rate of less than 1% as estimated by ProteinProphet. In addition, several proteins with lower initial probability were added to the list based on the manual validation of the corresponding peptide assignments to MS/MS spectra. In order to eliminate nonspecific binding proteins, the analysis of ENL-expressing cells was performed in parallel with control 293 cells. Those proteins that were also identified in the control gels were subtracted to create the final list of proteins that interact with ENL.

For the yeast 2-hybrid experiments, the baits and target plasmids were cotransfected into the yeast strain AH109 (Clontech) according to the manufacturer's instructions. GST pull-down was done with either purified GST or GST fused to full-length ENL and ³⁵S-labeled Dot1l produced by in vitro transcription/translation using TNT reagents from Promega (Madison, WI).

To test for histone methylation activity in immunoprecipitated material from 293fENL or HEK293 cells, EAP-loaded beads were incubated with 5 μg core histones (Upstate Biotechnology, Charlottesville, VA) and 1 μL ³H-methyl S-adenosylmethionine (3.7 × 10¹⁰ Bq/mmol [10 Ci/mmol], stabilized in H₂SO₄/EtOH; Hartmann Analytik, Braunschweig, Germany) in a total of 40 μL TE plus 1 mM DTT for 1 hour at 30°C. The reaction products were separated on 18% SDS-PAGE gels stained with Coomassie blue and treated with EnHance fluorographic solution (GE Healthcare) before exposure. For immunologic detection of in vitro methylated histones, the reaction was performed in the presence of 32 μM nonradioactive S-adenosylmethionine. After separation and transfer, the modified histones were detected by methylation-specific antibodies in an immunoblot procedure.

To detect pTEFb kinase activity in EAP, the proteins bound to ENL resin were washed once in 20 mM Tris-acetate (pH 7.9), 10 mM magnesium-acetate, 50 mM potassium-acetate, and 1 mM DTT. The beads were incubated with 1 μg recombinant GST-CTD (Calbiochem/Merck, Darmstadt, Germany) and 1 MBq of γ³²P ATP in the same buffer at 30°C for 1 hour. Phosphorylated proteins were visualized by SDS gelelectrophoresis, followed by Coomassie staining and autoradiography.

In vitro transcription was done with the HeLa scribe system (Promega) according to the instruction of the manufacturer. Inhibition of CDK9 was achieved by addition of 50 μM 5,6-dichloro-benzimidazole-riboside (DRB; SIGMA) as a 50-mM stock solution in DMSO. The run-off transcription was directed by a linearized DNA template containing the first 244 bases of the mouse Hoxa9 coding sequence under control of the cytomegalovirus (CMV) immediate early promoter.

ENL knock-down in HEK293 cells was achieved by serial transfection with pSIRENretroQ vectors capable of producing shRNA directed against ENL or luciferase as control. Cells were subjected at day 1 and again at day 4 to standard Ca transfection (20 μg of pSIRENretroQ vector per 6 × 10⁶ HEK293 cells) before analysis at day 7. For subsequent experiments, equal numbers of living cells as determined by trypan blue exclusion were applied.

Chromatin immunoprecipitation (ChIP) was performed according to the Upstate Biotechnology technical protocol as available at www.upstate.com. Global elongation rates were measured with nuclear run-on experiments according to Carey and Smale.²¹  In short, nuclei from 5 × 10⁷ cells were prepared by NP40 lysis (10 mM Tris/HCl [pH 7.4], 10 mM NaCl, 3 mM MgCl₂, and 0.5% NP40) and centrifugation. The nuclei were resuspended in 50 mM Tris/HCl (pH 8.3), 40% vol/vol glycerol, 5 mM MgCl₂, and 0.1 mM EDTA. Aliquots (26.5 μL) were mixed with 7 μL 5× run-on buffer (25 mM Tris/HCl [pH 8.0], 12.5 mM MgCl₂, 750 mM KCl, and 1.25 mM each of ATP, GTP, and CTP) and 1.5 μL of α³²P-UTP (1.11 × 10¹⁴ Bq/mmol [3000 Ci/mmol]). After incubation at 37°C for 15 minutes, RNA was isolated from the reaction mixtures by RNeasy column purification (Qiagen, Hilden, Germany). Purified RNA was spotted in dilutions onto nitrocellulose membranes and dried, and incorporated radioactivity was determined by PhosphoImager quantification (BioRad). For normalization, 18S stable RNA levels in the reaction mixtures were quantified by real-time reverse transcription–polymerase chain reaction (RT-PCR). Primers used were: 18S reverse, cttccttggatgtggtagccg; and 18S forward, tatcaactttcgatggtagtcgc.

The transformation capacity of MLL fusion proteins was determined by serial replating assays as previously reported.²²  In this experiment, transformation can be visualized as colony formation after repeated replating in semisolid medium. Replating assays were performed at least 3 times per construct.

For purification of EAPs, a stable producer cell line was established. In order to achieve low-level stable expression of flag-tagged ENL, the gene was introduced into HEK293 cells by transduction with a pMSCV-based retrovirus (pMSCV-fENL). Immunoblotting with anti-flag and anti-ENL antibodies and densitometric quantification demonstrated successful expression of fENL in 293fENL cells at approximately wild-type levels (Figure 1A). In order to verify that ectopic fENL is incorporated into a potential wild-type ENL complex, extracts of 293fENL cells were analyzed by gel filtration. Elution fractions were probed for the presence of flag-tagged proteins and ENL. In these experiments, the elution profile of ENL and fENL was nearly identical, and the majority of ENL/fENL protein was present in fractions corresponding to a relative molecular weight of approximately 400 kDa (Figure 1B). No uncomplexed fENL could be detected. Compared with the elution profile of wild-type ENL from untransduced HEK293 cells, the peak of flag-ENL elution was shifted slightly toward smaller molecular weights (Figure 1B). However, every elution fraction containing ENL in untransduced cells was also positive for flag-ENL in 293fENL cell extracts. This suggested that at least part of the ectopic fENL was incorporated into the appropriate endogenous complex or complexes.

EAPs were purified by a double immunoprecipitation strat-egy (Figure 2A). Triton/salt extracts from approximately 5 × 10⁸ 293fENL cells were precipitated with anti-flag agarose. After washing, bound material was selectively eluted with flag peptide and subjected to a second round of immunoprecipitation with a mixture of 3 bead-immobilized monoclonal antibodies targeted against ENL. Beads were washed again and residual material was released by low pH. For controls, the purification procedure was performed with extracts from nontransduced HEK293 cells, or mock precipitations were done with 293fENL cells using unliganded agarose. The final purified product was separated alongside controls by SDS-PAGE, followed by Coomassie and silver staining. This immunopurification reproducibly yielded a preparation of 9 major proteins detectable by Coomassie dye with approximately twice as many bands discernible after silver staining (Figure 2B). An identical pattern of EAPs was also obtained in the presence of nucleases (Figure 2C), excluding the possibility that EAP constituents were inadvertently copurifying due to an affinity for nucleic acids present in the cell extracts.

A total of 3 independent EAP preparations and the corresponding controls were subjected to mass spectrometry. The silver-stained gels were cut into slices, and proteins in each fragment were digested by trypsin. The resulting peptide mixtures were analyzed by reverse-phase LC-MS/MS using a linear ion trap mass spectrometer. In total, 15 proteins represented by more than 1 peptide could be unambiguously identified in the EAP samples (Table 1). All proteins (AF5q31, AF4, BCoR, and CBX8) that had been previously found in various 2-hybrid screens as interaction partners of ENL or the homologous AF9 were present also in the EAP preparations. In addition, EAP contained pTEFb, DOT1L, the AF4 homolog LAF4, and RING1, which is known to complex with CBX8, as well as the proteins BCoR, p52RO, TUB6, and PIP5K2C. Surprisingly, pTEFb was exclusively detected in the minor isoform as dimer of CDK9 with CYCT2, whereas approximately 80% of cellular pTEFb consists of CDK9 and CYCT1.²³  Also, HSP70, which is known to copurify with CDK9,²⁴  was part of EAP. Identities for individual silver-stained bands could be assigned for most proteins by the expected molecular weight and band-intensity/peptide-frequency correlations. Compared with ENL, the accompanying proteins were present in reproducible but nonequimolar stoichometries, implying that EAP is not a single complex but rather a mixture of different subcomplexes, with ENL as common unit. This assumption was corroborated by the fact that the sum of the calculated molecular weights of all identified EAP components exceeded 1.4 MDa, whereas ENL eluted during gel-filtration at approximately 400 kDa. Alternatively, it cannot be excluded that the conditions applied during size separation might have caused the disruption of a potentially larger complex.

To independently confirm the association of ENL with the major components of EAP, coimmunoprecipitations from unaltered HEK293 cells were performed (Figure 2D). To this end, Triton/salt extracts from native HEK293 cells were incubated either with ENL resin or with flag-agarose as control. The presence of DOT1L and all expected subunits of pTEFb (CDK9, CYCT2, and HSP70) could be successfully proven by immunoblotting. Interestingly, 2 CYCT2 immunoreactive bands were detected that corresponded exactly to the expected molecular weight of the known splice forms of CYCT2 (CYCT2B, 81 kDa; CYCT2A, 73.5 kDa). Mass spectrometry did not pick up the splice variants in purified EAP because no discriminatory peptides were present; however, both CYCT2 isoforms could be detected in EAP by immunoblotting (data not shown). Control blots did not detect any CYCT1 in EAP preparations (data not shown). In addition, the coprecipitation of endogenous ENL with AF4 also could be demonstrated. Suitable antibodies recognizing the AF4 homologs LAF4 and AF5q31 were not available.

The presence of DOT1L in EAP and the apparent important role for H3K79 methylation in MLL fusion protein mediated transformation^12,13  prompted us to analyze the DOT1L-ENL interaction more in detail. Because we were not able to obtain a full-length human DOT1L cDNA, the highly homologous mouse Dot1l was used in these experiments. Yeast 2-hybrid experiments were consistent with a direct association of the 2 full-length proteins (Figure 3A). A deletion analysis localized the ENL-binding domain within the Dot1l C-terminus between amino acids 826 and 1095, a region without any recognizable motif or homology. In contrast, the Dot1l interaction domain in ENL coincided with the highly conserved hydrophobic C-terminus that had been shown to be necessary and sufficient for transformation by MLL-ENL.²⁵  Interestingly, a small deletion of the ultimate 15 amino acids of ENL, a mutation that destroys ENL-dependent transactivation and MLL-ENL–mediated transformation, also precluded binding to Dot1l (Figure 3B,C). A direct interaction between ENL and Dot1l was also supported by GST pull-down experiments. In vitro transcribed and translated Dot1l could be coprecipitated with GST-ENL but not by GST alone (Figure 3D). Finally, EGFP-Dot1l and RFP-ENL fusions colocalized in the nuclei of living cells in a speckled distribution typical for ENL. An EGFP-Dot1l derivative with an internal 191–amino acid deletion within the ENL-binding domain was distributed exclusively in the cytoplasm, indicating an important role of this region for nuclear import und function of Dot1l (Figure 3E). ENL and Dot1l localized exclusively in the nucleus as demonstrated in cells counterstained with DAPI.

To test if the presence of DOT1L in EAP confers histone methyltransferase activity, purified EAP or a mock preparation obtained by precipitation from 293fENL cells with unliganded agarose beads were incubated with core histones in the presence of ³H-methyl-S-adenosylmethionine. The reaction mixture was separated by SDS-PAGE, stained with Coomassie, and treated with a fluorographic enhancer to allow visualization of ³H-labeled compounds. This assay detected a robust histone H3–specific methyltransferase activity associated with EAP (Figure 4A). In concordance with the known substrate specificity of DOT1L, free recombinant histone H3 was not a substrate for methyl-transfer by EAP (not shown). A specific increase of histone H3 lysine 79 dimethylation could be seen in core histone preparations after incubation with EAP and S-adenosylmethionine, but not in controls. Since DOT1L is the only known methyltransferase able to catalyze H3K79 dimethylation,²⁶  it is highly likely that DOT1L is responsible for the histone-methylating activity in EAP.

In an analogous approach, EAP was tested for pTEFb kinase activity in vitro. For this purpose, purified recombinant GST– RNA polymerase II C-terminal domain kinase (RNAPolII CTD) was incubated with EAP or a mock control precipitate in a reaction containing γ³²P-ATP. Autoradiography revealed a robust kinase activity of EAP (Figure 4B). The influence of EAP on in vitro transcription was assessed by run-off transcription from a template containing the 5′ portion of the Hoxa9 gene driven by the CMV immediate early enhancer. For detection of transcripts, the reaction was done in the presence of ³²P-UTP. As expected, addition of the pTEFb inhibitor DRB caused a severe reduction in transcript levels. However, supplementation of the transcription reactions with EAP partially relieved the inhibitory effect of DRB, suggesting that EAP supplies additional pTEFb activity (Figure 4B).

The impact of EAP inhibition on global H3K79 methylation was examined by knock-down of ENL. For this purpose, HEK293 cells were transfected with constructs encoding shRNAs directed against ENL or luciferase as control. shRNAs are processed by cellular enzymes into inhibitory RNAs that trigger the destruction of their complementary mRNAs. The efficiency of ENL knock-down was controlled at the protein level by immunoblotting and densitometric evaluation. Whereas stable knock-down clones with reduced ENL levels could not be obtained (data not shown), transient shRNA transfection reduced endogenous ENL concentrations to approximately 20% of controls after 7 days and did not cause increased cell death compared with a luciferase knock-down control (Figure 5A; data not shown). Correspondingly, global H3K79 dimethylation also decreased to approximately 30% compared with the value obtained in control shRNA–treated cells (Figure 5A). Since it has been shown by independent laboratories that DOT1L is directly responsible for H3K79 dimethylation in HEK293 cells,^26,27  it is highly likely that perturbation of EAP and therefore DOT1L function was the cause for the reduced modification levels. The transient knock-down of ENL did not significantly affect expression of pTEFb components in the cells (data not shown).

The distribution of dimethyl-H3K79 was determined in ENL knock-down cells across the HOXA9 coding region by ChIP. HOXA9 is expressed in HEK293 cells and it is known that HOXA genes are particularly sensitive to a loss of H2B ubiquitinylation and therefore also H3K79 methylation because the former is a prerequisite for the latter to occur.²⁸  Gene-specific H3K79 dimethylation was moderately but significantly reduced after shRNA-mediated knock-down of ENL (Figure 5B). In contrast, H3K9 dimethylation did not change significantly across the HOXA9 coding region after ENL knock-down. Also, H3K4 di- and trimethylation was largely unaffected except for a localized reduction around the start of the coding sequence. Overexpression of ENL did not increase global or local H3K79 methylation, indicating that ENL is not limiting under normal cellular conditions (data not shown).

Because DOT1L and pTEFb are enzymatic activities associated with elongation, we wanted to determine if a loss of EAP function affects transcription at this level. For that purpose, transcriptional elongation was quantified using a nuclear run-on assay, which measures incorporation of radioactive nucleotides into nascent mRNAs produced in isolated nuclei. Under these conditions, new initiation generally does not occur.²¹  Therefore, radioactive labeling is proportional to elongation. Nuclei were prepared from HEK293 cells depleted for ENL by shRNA and from control cells transfected with luciferase shRNA. Identical numbers of nuclei were incubated with ³²P-αUTP. RNA was purified by spin columns and spotted on membranes, and radioactivity was quantified by phosphoimaging. As a normalization control, the concentration of stable 18S RNA was quantified by real-time RT-PCR. An approximately 30% reduction in global elongation rates was detectable in ENL knock-down cells compared with controls. To determine if the consequences of ENL knock-down are gene specific, labeled run-on RNA was hybridized to select cDNAs. As expected, the effect on elongation varied widely depending on the gene examined. Whereas run-on rates were reduced approximately 40% for AF4, 20% for AF5q31, and 10% for ENL, HOXA10 elongation was not sensitive to reduced ENL levels. Elongation rates for HOXA9 were too low to be detected by a run-on/hybridization assay; however, a knockdown of ENL caused a significant reduction of total HOXA9 mRNA abundance as determined by q-RTPCR (Figure 5C). A dose dependent inhibitory effect of ENL-specific shRNA was also observed in transient luciferase reporter experiments testing 3 different promoters of viral (simian virus 40 [SV40], CMV) and cellular origin (Hoxa7; Figure 5D). In summary, these results underscore an important role for EAP in general transcription in vivo also.

The 15 C-terminal amino acids of ENL are essential for binding of ENL to DOT1L. The same region also is responsible for a direct interaction between ENL and CBX8. In addition, this peptide has been shown to be absolutely required for the transforming ability of the corresponding MLL-ENL fusion protein. The importance of the ENL-DOT1L interaction for the transforming capacity of MLL-ENL was examined by an alanine-scanning mutagenesis. A total of 2 different amino acid triplets within the ENL C-terminus were replaced by alanine. The mutant ENL sequences were introduced into MLL-ENL and into a 2-hybrid bait construct (Figure 6A). Correct expression was demonstrated by immunoblotting of the respective GAL4-DNABD and MLL fusion proteins. Importantly, the transforming capability of the MLL-ENL mutants was highly correlated with the binding affinity for DOT1L. Only ENL mutant A, which bound efficiently to DOT1L, was transforming in the context of MLL-ENL. In contrast, mutant B had significantly reduced affinity for DOT1L and failed to transform when fused to MLL. No such relationship, however, was detected for the interaction of ENL with CBX8, as this parameter was comparable for both mutants (Figure 6B-C). The transforming potential of the individual MLL fusion proteins was perfectly correlated with their ability to induce the expression of the known MLL-ENL target genes Hoxa7, Hoxa9, and Meis1 in primary hematopoietic cells (Figure 6D).

A still largely unsolved question of MLL fusion biology pertains to the biological activity contributed by the fusion partner. The results presented here help to answer this problem by suggesting a function for 1 of the most common nuclear MLL fusion partners, the ENL protein. ENL associates with a set of proteins that can be purified from cells as macromolecular assembly. The composition of EAP as determined here is backed up by a number of previous results, mainly from 2-hybrid screens that indicated a direct interaction of ENL and/or the homologous AF9 fusion partner with AF5q31 and AF4^10,11  as well as with CBX8 and BCoR.^14,16  The fact that AF10 was not reproducibly found in EAP confirms our earlier results that a direct interaction of ENL with AF10 occurs only if AF10 is truncated.¹¹  The internal ENL binding site probably is not accessible under regular cellular conditions. In a similar direction, AF9 should not be a major constituent of EAP because ENL and AF9 occupy the identical binding site in AF4 and AF5q31, respectively.^10,11  Therefore, our purification strategy that selects for ENL-bound AF4/AF5q31 should not simultaneously also bring down AF9. Most probably the EAP proteins are not present in a single protein complex. Structure function studies^10,11  have shown that AF5q31, AF4, and possibly also the highly homologous LAF4 interact with the same C-terminal domain of ENL. It is very unlikely that 3 proteins with individual molecular weights of more than 130 kDa could simultaneously occupy the identical binding surface on the much smaller ENL protein. Likewise, CBX8 and, as shown in this report, DOT1L share the same interaction domain in ENL, and therefore binding is expected to be mutually exclusive. Considering the stoichometry of the EAP constituents, we propose that ENL and pTEFb are core components that associate depending on context with a select set of the other EAP proteins. Although it cannot be excluded that gel filtration itself might disrupt a larger protein complex, the presence of more than 1 complex that shares ENL as a common subunit would more likely explain the discrepancy of the apparent molecular weight of EAP in gel filtration compared with immunoprecipitation.

The presence of both pTEFb and DOT1L, 2 enzymatic activities necessary for transcriptional elongation, suggests a function for the EAP assembly during this process. pTEFb is a dimer of the cyclin-dependent kinase 9 that, assisted by heat-shock proteins, associates with either cyclin T1 or cyclin T2 to form an active kinase that phosphorylates serine 2 in the heptad repeat of RNA polymerase II.^24,29  This modification is necessary for productive elongation during the transcription of many genes. The connection of pTEFb with MLL partner proteins is also supported by a report that identified AF5q31 in affinity-purified pTEFb preparations.³⁰  DOT1L is a non-SET methyltransferase with specificity for histone H3 lysine79 and is the only known enzyme with this histone specificity.²⁶  Previously, we demonstrated that ENL interacts with histone H3 via N-terminal sequences.¹¹  Here, we show ENL also interacts directly with DOT1L. ENL is therefore a prime candidate for serving as an adaptor that tethers DOT1L to its substrate. During the preparation of this manuscript, a report became available that lends further support to a role for EAP in elongation control. Although done exclusively with transiently overexpressed proteins, Bitoun et al³¹  demonstrated that AF4 associates with AF9, ENL, and pTEFb to stimulate transcriptional elongation through a recruitment of DOT1L and stimulation of H3K79 dimethylation. The same group showed previously that the “robotic” mouse harbors a mutant of AF4 that impairs proteasomal degradation of this protein.³²  Robotic mice therefore overexpress AF4, and this overexpression is also accompanied by higher levels of H3K79 dimethylation in the cerebellum, the organ responsible for the “robotic” phenotype. In these studies, CDK9 was expressed in combination with CYCT1, which most likely does not match the endogenous EAP composition, because we could not detect any CYCT1 in our preparations. To our knowledge, EAP is the first instance where a combination of CDK9 with CYCT2a/CYCT2b forms the active unit. EAP components are ubiquitously expressed, and a central role for ENL in cellular biochemistry is supported by genetic ablation experiments in which ENL-deficient embryos die extremely early in utero, even before implantation.³³  In summary, all available evidence points clearly to a function of EAP as a major control instance for transcriptional elongation, and therefore EAP seems to be essential for efficient gene expression in all cells.

Surprisingly, EAP also contained minor quantities of 3 proteins (CBX8, RING1, and BCoR) involved in polycomb-mediated repression. The chromoboxprotein CBX8 (also known as human polycomb 3 or HPC3) and RING1 are common components of the polycomb repressive complex 1.³⁴  BCoR has been previously detected in a corepressor complex recruited by the BCL-6 protein.³⁵  Both CBX8 and BCoR have been identified in interaction screens as direct binding partners of ENL or AF9, arguing against an artefactual copurification.^14,–16  It is therefore tempting to speculate that EAP might also exist in a repressive mode. Indeed, there are reports that a complex of the ENL homolog AF9 and DOT1L is involved in aldosterone-controlled repression of a sodium channel gene in renal cells.³⁶ 

The identification of EAP also has implications for our understanding of the transforming mechanism of the MLL-ENL fusion protein. Within the context of the leukemogenic MLL-ENL, the biological activities of EAP would be constitutively mistargeted via the fusion partner. As a consequence, elongation and therefore gene expression would be inappropriately stimulated. Indeed, a dramatic increase of H3K79 dimethylation is a hallmark of target genes activated by MLL-ENL.¹³  This correlates well with our results that mutations in ENL that impair the interaction with DOT1L also inactivate the transforming potential of the corresponding MLL-ENL fusion. Interestingly, the unrelated fusion MLL-AF10 also works through recruitment of DOT1L.¹²  It remains to be seen if chromatin modification and transcriptional elongation are important also for other MLL fusion proteins. If this is the case, drugs that potentially interfere with these mechanisms might be attractive therapeutic agents.

We thank Renate Zimmermann and Gaby Sander for technical support and B.C. Kone for sharing reagents.

This work was supported by DFG grants SL27/6–2 and SFB473/D2 (R.K.S.) and National Institutes of Health grants CA78815 and CA92251 (J.L.H.). M.-P.G.-C. acknowledges funding in part by Dr von Haunersches Kinderspital, Munich. Equipment funding from Jose-Carreras-Stiftung and Curt-Bohnewand-Fond, and travel support by BaCaTec is gratefully acknowledged.

National Institutes of Health

Contribution: R.K.S. designed research; D.M., C.B., D.Z., M.-P.G.-C., S.M., A.S., R.Z., A.N., and A.C. performed research and collected data; A.C., J.L.H., and R.K.S. analyzed and interpreted data; and J.L.H. and R.K.S. drafted the manuscript.

D.M. and C.B. contributed equally to this work.

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

Correspondence: Robert K. Slany, Genetics, University Erlangen, Staudtstrasse 5, 91058 Erlangen, Germany; e-mail:rslany@biologie.uni-erlangen.de.