|Home | About | Journals | Submit | Contact Us | Français|
The mixed-lineage leukemia gene (MLL, ALL1, HRX) encodes a 3,969-amino-acid nuclear protein homologous to Drosophila trithorax and is required to maintain proper Hox gene expression. Chromosome translocations in human leukemia disrupt MLL (11q23), generating chimeric proteins between the N terminus of MLL and multiple translocation partners. Here we report that MLL is normally cleaved at two conserved sites (D/GADD and D/GVDD) and that mutation of these sites abolishes the proteolysis. MLL cleavage generates N-terminal p320 (N320) and C-terminal p180 (C180) fragments, which form a stable complex that localizes to a subnuclear compartment. The FYRN domain of N320 directly interacts with the FYRC and SET domains of C180. Disrupting the interaction between N320 and C180 leads to a marked decrease in the level of N320 and a redistribution of C180 to a diffuse nuclear pattern. These data suggest a model in which a dynamic post-cleavage association confers stability to N320 and correct nuclear sublocalization of the complex, to control the availability of N320 for target genes. This predicts that MLL fusion proteins of leukemia which would lose the ability to complex with C180 have their stability conferred instead by the fusion partners, thus providing one mechanism for altered target gene expression.
homeotic genes of the invertebrates and Hox genes of the vertebrates encode a family of related transcription-regulatory proteins that play critical roles in implementing developmental patterns and conferring positional information (18). Development of a normal body plan requires continued expression of each homeotic/Hox gene throughout development, and the combination of their expression patterns determines the segmental identity of cells throughout the body. Deregulation of these expression patterns causes transformation of segmental identities during development and can contribute to a variety of human disorders, including malignancy. Hox gene expression is initiated by transiently expressed segmentation gene products. However, the maintenance of appropriate Hox expression is conferred by two classes of regulators, the positive regulatory trithorax group (trx-G) and the negative regulatory polycomb group (Pc-G), which apparently function through epigenetic mechanisms (21, 26).
Mll is a mammalian member of the trx-G, and loss-of-function studies indicate that it is required for midgestational development. An Mll-exon3lacZ homozygous disruption results in embryonic death at E10.5 with multiple patterning defects including abnormal ganglia, absence of the maxillary branch of the first branchial arch, and pooling of erythroid precursors in the coelomic cavity (35). This Mll knockout revealed that Mll is required for the maintenance rather than the initiation of Hox gene expression in early embryogenesis (34). Although Mll-null embryos demonstrated correct initiation of Hox gene expression, including Hoxa7 and Hoxc8, the subsequent maintenance of the expression of these Hox genes was lost. Together, these studies indicate that Mll is critical for the proper maintenance of Hox gene expression during development.
Translocations of human chromosome segment 11q23 disrupt the MLL gene, resulting in a common chromosome abnormality encountered in acute leukemia, which is found in 80% of infant leukemias and the majority of chemotherapy-related leukemias following treatment with topoisomerase II inhibitors (12, 30). The MLL gene is composed of 37 exons and encodes a 3,969-amino-acid (aa) nuclear protein with multiple distinct domains including an N terminus with three AT-hook motifs, a DNA methyltransferase homology domain (DNMT), four PHD fingers, a transactivation domain, and a highly conserved SET domain at the C terminus (11, 15, 31) (see Fig. Fig.1A).1A). MLL is fused with more than 20 different chromosomal partners on translocation, resulting in multiple unique MLL fusion proteins. The MLL breakpoints are tightly clustered in an 8.3-kb genomic region containing exons 8 to 14, and the derivative 11 chromosome [der(11)] encodes the MLL fusion protein, which possesses ~1,300 aa of N-terminal MLL (12). The C-terminal fusion partners of MLL are very diverse, ranging from putative transcription factors to cytoplasmic structural proteins. This has raised the question whether there is any common property conferred by the fusion proteins (2).
In this study, we demonstrate that native MLL is proteolytically processed into N-terminal p320 (N320) and C-terminal p180 (C180) fragments. We identify two protease recognition motifs that are highly conserved in MLL homologues from Drosophila to mammals. Processed MLL fragments form a stable complex based on interactions between the FYRN domain of N320 and the FYRC and SET domains of C180. Mutational analysis indicates that interaction between C180 and N320 is required for steady-state levels of N320 and correct subnuclear localization. These observations indicate that leukemogenic MLL fusion proteins would lose association with C180. Furthermore, they suggest a model in which the varied MLL fusion partners all play a common role of stabilizing the fusion product.
Flag-tagged full-length human MLL cDNA inserted in the pCXN2 vector has been described (16). Deletion between two SpeI sites generates MLLΔ2253-2727. Oligonucleotides that encode the Flag or Myc epitope were inserted into the unique BstEII site of MLL to generate C-terminally Flag- or Myc-tagged MLL, respectively. Flag-tagged truncated MLL constructs were generated by inserting a stop codon into a SalI or PflMI site. A PCR fragment encoding aa 2720 to 3969 was inserted into the pCMV-Myc vector (Clontech) to generate Myc-tagged C180. A stop codon was added at the BstEII site to generate Myc-C150ΔF/S. Green fluorescent protein (GFP) was inserted in frame at the BstEII site of MLL. A PCR fragment encoding red fluorescent protein (RFP) was cloned into pCDNA3, and the resultant vector was used to generate the RFP-MLL-GFP construct. Truncation mutants with a deletion of FYRC and SET domains (ΔF/S) were constructed by adding a stop codon after either the Flag epitope or GFP. A NotI site was created in wild-type (wt) MLL to generate DG→AA and DGV→AAA MLL mutants.
MLL fragments consisting of aa 5 to 955 and aa 2829 to 2883 were expressed as glutathione S-transferase GST fusions in Escherichia coli BL21/DE3. GST-MLL-fusion proteins were purified as previously described (13) and injected into rabbits (Covance). Abs were purified from immunized rabbit serum by using a HiTrap protein A column (Amersham Pharmacia Biotech) as specified by the manufacturer.
Human embryonic kidney 293T cells were transiently transfected with the indicated MLL-expressing constructs by using Fugene 6 (Roche Molecular Biochemicals) as specified by the manufacturer. Cells were solubilized in RIPA buffer 48 h after transfection. Immunoprecipitation and immunoblot assays were performed as described previously (6) using indicated anti-Flag (Sigma), anti-N-terminal (anti-NT), or anti-CT Ab.35S-labeled MLL polypeptides were generated by in vitro transcription and translation using the TNT-coupled reticulocyte lysate system (Promega). Immunoprecipitation assays of35S-labeled polypeptides using anti-Flag antibody were described previously (6).
Fluorescence-tagged MLL was expressed in 293T cells. Cells were fixed in phosphate-buffered saline with 3% paraformaldehyde, stained with YOPRO-3 (Molecular Probes), and detected by confocal microscopy.
An MLL construct carrying a Flag tag at the C terminus was transfected into 293T cells. Cellular extracts purified from 30 10-cm dishes were incubated with anti-Flag Ab-coated beads. Immunoprecipitates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a polyvinylidene difluoride (PVDF) membrane. Coomassie blue staining demonstrated one predominant band, which was excised and subjected to N-terminal Edman degradation analysis.
To further explore the function of MLL and MLL fusion proteins, we generated several Abs against distinct regions of MLL, all of which failed to detect a protein appropriate for full-length MLL (predicted molecular mass, 430 kDa; actually migrates at ~500 kDa). However, an Ab to an N-terminal region (aa 5 to 955; anti-NT) recognized instead an ~320-kDa fragment in cells expressing a full-length MLL construct (Fig. (Fig.1C).1C). Reciprocally, an Ab to a C-terminal region (aa 2829 to 2883; anti-CT) also failed to detect full-length MLL but did identify an ~180-kDa fragment (Fig. (Fig.1D).1D). This suggested that MLL may undergo a posttranslational cleavage. To test this hypothesis, we tagged the N-terminal or C-terminal portion of MLL with a Flag epitope (Fig. (Fig.1B).1B). An anti-Flag Ab detected the same 320- and 180-kDa fragments (Fig. (Fig.1E),1E), as did the anti-NT (Fig. (Fig.1C)1C) and anti-CT (Fig. (Fig.1D)1D) Abs. This confirmed that full-length MLL is cleaved into N-terminal p320 (N320) and C-terminal p180 (C180) fragments. A series of internal deletion mutants were subsequently generated to localize the cleavage site. The Flag-tagged mutant (MLLΔ2253-2727) was not cleaved and was detected as a ~450-kDa protein by all three Abs (Fig. (Fig.1B1B to E). In total, these data indicate that MLL is cleaved and that the cleavage site(s) resides between aa 2253 and 2727.
Numerous studies have demonstrated a punctate subnuclear localization for MLL utilizing Abs raised against the N-terminal portion of the protein (4, 5, 24, 32). Since MLL is proteolytically processed to two distinct fragments, one critical issue is whether C180 demonstrates the same or a unique subcellular localization compared to N320. To address this, RFP was fused to the N terminus of MLL and GFP was introduced into the C terminus. When this doubly tagged protein was expressed, it was cleaved correctly into RFP-N320 and GFP-C180 fragments (Fig. (Fig.2D).2D). Expression of this dually fluorescent MLL protein in 293T cells resulted in the colocalization of RFP-N320 and GFP-C180 in a punctate nuclear pattern, as determined by confocal microscopy (Fig. (Fig.2A).2A). This observation raised the possibility of a physical association between N320 and C180 fragments prompting the creation of a dually tagged construct bearing a Flag-tagged N terminus and a Myc-tagged C-terminus (Fig. (Fig.2B).2B). Myc-C180 was coimmunoprecipitated with Flag-N320 by using the anti-Flag Ab (Fig. (Fig.2B).2B). A reciprocal immunoprecipitation using an anti-Myc Ab coprecipitated the Flag-N320 as well as C180 fragments (data not shown). To examine whether endogenous MLL protein was proteolytically cleaved and assembled, cellular extracts from HeLa and 293T cells were immunoprecipitated with anti-CT Ab and subjected to Western blot analysis using either anti-CT or anti-NT Abs. The endogenous MLL protein of both cells was also processed into N320 and C180 fragments with evidence through their coimmunoprecipitation of a post-cleavage complex (Fig. (Fig.2C2C).
To precisely determine the cleavage site, the Flag-tagged C180 MLL fragment was purified from immunoprecipitates separated by SDS-PAGE and transferred to a PVDF membrane. A Coomassie blue staining step identified one predominant 180-kDa peptide (Fig. (Fig.3A).3A). N-terminal Edman degradation analysis of the excised polypeptide yielded a 12-aa sequence corresponding to aa 2719 to 2730 of MLL (Fig. (Fig.3A).3A). Sequence alignment of MLL from different species indicated that this cleavage site is highly conserved and resides within the internally deleted region of the Drosophila trxE3 mutant (20). However, mutation of the cleavage site from D/GVDD to A/AADD (Fig. (Fig.3B)3B) reduced but did not completely abolish the N320 band; however, it did reveal an increase in the amount of apparently full-length p500 MLL. Of note, this mutant generated a C-terminal fragment somewhat larger than wild-type C180 (Fig. (Fig.3B).3B). This compilation of findings suggested the existence of an additional cleavage site upstream of the residue 2718-2719 site. A search for a similar cleavage motif identified another potential proteolytic site, D/GADD, located 53 residues proximal to the initially identified D/GVDD site. An MLL construct possessing doubly mutated motifs (Fig. (Fig.3B)3B) was no longer processed, as evidenced by the disappearance of both N320 and C180 fragments as well as an abundance of full-length p500 (Fig. (Fig.3B).3B). Thus, MLL appears to be proteolytically processed at two independent sites, which we have called CS1 (D/GADD, aa 2666-2667) and CS2 (D/GVDD, aa 2718-2719).
The interaction sites between N320 and C180 were further refined by coexpressing truncations of the N terminus along with C180 (Fig. (Fig.4A).4A). The region between aa 1394 and 2160 within the N320 fragment proved necessary for its interaction with C180 since residues 1 to 2160 but not residues 1 to 1393 would coimmunoprecipitate Myc-C180 (Fig. (Fig.4E).4E). Several conserved domains lie within this 767-aa region, including PHD fingers 1 to 3, a Bromo domain, PHD finger 4, and a FYRN domain (Fig. (Fig.1A).1A). Three conserved regions have been recognized within the C180 fragment of MLL. They include the transactivation domain that mediates CBP interaction, the SET domain at the C-terminal end, and the FYRC domain immediately N-terminal to the SET domain (Fig. (Fig.1A).1A). We reasoned that interaction between N320 and C180 is likely to be mediated through conserved domains. To test this hypothesis, we performed a series of coimmunoprecipitation assays with the indicated MLL domains (Fig. (Fig.5).5). These analyses identified a direct interaction between the FYRN domain of N320 MLL and the region containing both the FYRC and SET domains of C180 MLL (Fig. (Fig.5G5G).
To investigate the functional significance of the N320-C180 interaction mediated by FYRC plus SET, we deleted C-terminal residues 3621 to 3969 of MLL, which contain both domains (MLLΔF/S, Fig. Fig.6).6). This C-terminally truncated mutant, MLLΔF/S, was proteolytically processed to the expected N320 fragment and the predictably smaller C150ΔF/S fragment (Fig. (Fig.6).6). The C150ΔF/S fragment did not coprecipitate N320 (Fig. (Fig.6).6). This confirms the in vitro interaction between the FYRC-plus-SET region of C180 and N320. Of note, although the C150ΔF/S fragment was of comparable abundance to wt C180, we observed a 10-fold reduction in N320 levels in cells with mutant C150ΔF/S (Fig. (Fig.6).6). This suggests that the interaction between N320 and C180 protects N320 from degradation. To further demonstrate the effect of the N320-C180 interaction on N320 stability, we coexpressed a Flag-tagged N-terminal MLL with either Myc-tagged C180 or C150ΔF/S in 293T cells (Fig. (Fig.7A).7A). In the presence of C180, the N-terminal MLL was stabilized and detected at high abundance (Fig. (Fig.7B,7B, lane 1) compared to the markedly diminished level detected when expressed with C150ΔF/S (lane 2). To estimate the half-life of the N-terminal MLL fragment, cells expressing these constructs had their protein synthesis inhibited by cycloheximide. Levels of N-terminal MLL were determined over a serial time course by anti-Flag Western blot analyses. In the presence of C180, the estimated half-life of N-terminal MLL was longer than 8 h (Fig. (Fig.7C).7C). In contrast, when C150ΔF/S, which fails to interact, was expressed, the estimated half-life of N-terminal MLL was shortened to ~4 h (Fig. (Fig.7D).7D). This confirms that the processed MLL N-terminal fragment must interact with the C-terminal fragment to avoid degradation and confer stability. Thus, one rationale for a dynamic interaction between cleaved MLL fragments is to regulate the levels of the N-terminal fragment of MLL.
To assess the effects of interaction on subcellular protein distribution, the MLLΔF/S construct lacking the FYRC and SET domains was labeled with RFP at its N terminus and GFP at its C terminus (Fig. (Fig.8).8). Confocal microscopy of cells expressing this doubly fluorescent construct revealed that there was no detectable expression of the RFP-N320 fragment (Fig. (Fig.8,8, left upper panel). Thus, because of the low level of RFP-N320, we could not accurately assess whether N320, which possesses previously defined subnuclear localization sequences (32), was itself capable of correct localization in the absence of interaction with C180. In contrast, GFP-C150ΔF/S was easily detected in the nucleus but had lost its normal punctate subnuclear pattern and was now distributed diffusely throughout the nucleus (Fig. (Fig.8).8). This argues that the localization of C180 to the subnuclear compartment is governed by its association with N320, although we cannot exclude the possibility that localization is mediated directly through the FYRC and SET domains. Thus, the processing of MLL regulates both the abundance and the localization of MLL fragments, providing mechanisms to control target gene expression.
Specific proteolytic cleavage plays a critical role in many biological pathways, including the liberation of SREBP from the endoplasmic reticulum, the release of Notch from the plasma membrane, the maturation of hedgehog signaling molecules, and the activation of caspases and their subsequent cleavage of death substrates during apoptosis. Here, we demonstrate that mammalian MLL is proteolytically processed at cleavage sites which follow P1 aspartic acid residues located within two highly conserved D/GVDD and D/GADD motifs. The precise protease which mediates this cleavage remains to be determined but is likely to be the same for the similar CS1 and CS2 motifs. Of note, we have tested inhibitors of calpain, the 26S proteasome, and caspases and demonstrated no substantial effect on the cleavage. We also noted a D/GVDD motif deleted in a Drosophila TRX that is located within a 281-aa region deleted in a described mutant trx allele, trxE3 (32). Unlike null mutations of trx that fail to maintain the expression of homeotic genes of the bithorax (BX-C) and antennapedia (ANT-C) complexes (3), trxE3 displays selectively reduced expression of ANT-C but not BX-C genes (28). Subsequent studies using antibodies recognizing different regions of trx are consistent with the notion that this recognized motif serves as a proteolytic site (19). The homeotic defects in trxE3 argue that cleavage of trx has a developmental consequence.
The precise mechanisms whereby MLL exerts its principal function to maintain the expression of target genes, including Hox genes, remain uncertain. However, the presence of conserved motifs with MLL provides some insight. MLL possesses domains with ascribed biochemical properties: the three AT hook motifs provide minor-groove DNA binding (31, 36), the DNMT homology domain binds methylated DNA differentially and can repress transcription (8, 11, 22, 36), a conserved subnuclear localization signal directs punctate nuclear distribution (32), the PHD fingers mediate protein-protein interaction (14), the transactivation domain interacts with CREB-binding protein (13), and the highly conserved SET domain binds to histone H3, Sbf1, and ASH1, as well as itself (2, 9, 17, 25). Here we identify the D/GVDD and D/GADD motifs and assign a role for them in MLL processing as well as identifying roles for the FYRN and FYRC domains. The FYRN and the FYRC plus SET domains mediate the interaction between the N320 and C180 subunits of MLL. Of note, Drosophila TRX also possesses FYRN and FYRC motifs, suggesting that it may also form a complex of processed fragments. Two distant MLL family members, ALR and MLL3, are present in the human genome (23, 29) but lack the D/GVDD cleavage motifs. Interestingly, they have a varied organization of domains compared to MLL, possessing a PHD domain, a FYRN domain, a FYRC domain, and a SET domain all clustered within the C-terminal 450 aa of ALR and MLL3. This highly conserved architecture is also present in Caenorhabditis elegans and Drosophila ALR homologues (27). The interaction between the FYRN and the FYRC plus SET domains noted here for MLL raises a testable possibility that ALR and MLL3 may have clustered these domains to create a similar functional moiety. Perhaps one role for the cleavage of MLL is to physically juxtapose these previously separated domains.
Chromosomal translocations in human leukemia which disrupt MLL produce a chimeric protein in which the N-terminal portion of MLL is fused to 1 of more than 20 partners. These fusion partners do not have any recognizable motif in common, and they range widely from nuclear factors to cytoplasmic proteins and cell junction proteins. Gene expression profiles of human leukemias bearing an MLL translocation identified a pattern of upregulated genes. Among these genes were some of the well-recognized targets of wt MLL, including HOXA4, HOXA5, and HOXA9 (1, 33). This argues that the N-terminal 1,300-aa portion of MLL is sufficient to confer at least some target gene specificity to MLL fusion proteins. Mice expressing an MLL-AF9 fusion product progressed to acute myeloid leukemia, establishing the oncogenic potential of MLL chimeric proteins (7). In contrast, Mll-exon8Myc mice in which a small myc epitope tag was fused with MLL failed to develop leukemia, arguing that a truncated MLL fragment was not in itself sufficient to cause transformation. However, Mll-exon8LacZ mice in which MLL was fused with the β-galactosidase protein did progress to acute leukemia (10). β-Galactosidase has been extensively used in transgenic mouse models and plays no independent, demonstrated role in tumor formation. The compilation of these findings indicates that individual fusion partners play an indispensable role but that they need not be directly oncogenic. The observations here concerning protein instability may provide an explanation. The FYRN domain, required for interaction of the processed MLL fragments, is located downstream of the chromosomal breakpoint, indicating that MLL fusion proteins would lose interaction with the stabilizing C180 component. Consequently, one testable characteristic that might prove to be shared by all the partner proteins is the capacity to confer protein stability to the MLL N terminus.
This work was supported in part by NIH award P01 CA68484. P.E. is a recipient of the Beckenstein Fellowship and a Special Fellow of the Leukemia and Lymphoma Society.
We thank E. Smith for editorial assistance and B. Malynn for helpful discussions and suggestions.