Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis

Abstract

MLL1 fusion proteins activate HoxA9 gene expression and cause aggressive leukemias that respond poorly to treatment, but how they recognize and stably bind to HoxA9 is not clearly understood. In a systematic analysis of MLL1 domain recruitment activity, we identified an essential MLL1 recruitment domain that includes the CXXC domain and PHD fingers and is controlled by direct interactions with the PAF elongation complex and H3K4Me2/3. MLL1 fusion proteins lack the PHD fingers and require prebinding of a wild-type MLL1 complex and CXXC domain recognition of DNA for stable HoxA9 association. Together, these results suggest that specific recruitment of MLL1 requires multiple interactions and is a precondition for stable recruitment of MLL1 fusion proteins to HoxA9 in leukemogenesis. Since wild-type MLL1 and oncogenic MLL1 fusion proteins have overlapping yet distinct recruitment mechanisms, this creates a window of opportunity that could be exploited for the development of targeted therapies.

Figure 1. Recruitment of the MLL1 complex is required for HoxA9 activation

(A) Important MLL1 domains and protein interactions. The MLL1 protein is proteolytically cleaved in the cell into N terminal (MLL1N) and C terminal (MLL1C) proteins. All MLL1 leukemogenic fusions break just before the first PHD finger (line, Leukemia Breakpoint). Two MLL1 fusion proteins MLL-AF9 and MLL-ENL are shown as examples. (B) HoxA9 protein expression depends on the expression of MLL1. Extracts were prepared from Mll1−/− mouse embryonic fibroblast (MEF) knockout cells (lane −) and Mll1−/− MEF cells that have been reconstituted with full length MLL1 (lane +). Western blots were probed with the antibodies indicated. (C) Chromatin immunoprecipitation (ChIP) experiments at HoxA9 in Mll1−/− (blue line) versus Mll1−/− reconstituted with MLL1 (red line) MEF cells using the antibodies indicated. A schematic of the mouse HoxA9 locus with red boxes indicating the positions of 6 primer/probe sets is shown at the bottom of the ChIP figures. ChIP results shown are typical for at least two independent experiments. Error bars represent standard deviation of three PCR reactions. See also Figure S1.

Figure 2. The CXXC domain and PHD fingers are sufficient for targeting MLL1 to the HoxA9 locus in vivo

(A) Legend for the ChIP experiments in C. Blue line= ChIP in cells transfected with an empty vector, Red line= ChIP in cells transfected with the constructs indicated and Grey dots = the binding pattern of full length MLLN for comparison. (B) Schematics showing the series of deletion constructs. (C) The regions containing the MLL1 CXXC domain and the PHD fingers are both necessary for targeting MLL1 to the HoxA9 locus. Different fragments of the MLL1 N terminus were FLAG and HA (fh) double tagged and expressed in Mll1−/− MEFs and α-HA ChIP experiments were performed. ChIP results shown are typical for at least three independent experiments. Error bars represent standard deviation of three separate PCR reactions. See also Figure S2.

Figure 3. The MLL1 complex interacts specifically with the PAF elongation complex

(A) 293 cells stably expressing either FLAG-HA tagged (fh) CXXC:1067–1432 or fh-P1-P4:1432–1989 were subjected to a one step FLAG purification. Mass spectroscopy analysis identified components of the MLL1 core complex (HCFC1 and HCFC2) and components of PAF1C (shown in red) (B) FLAG IP’s were done from 293 cell nuclear extracts expressing either an empty vector (1), fh-CXXC:1067–1432 (2) or fh- P1-P4:1432–1989 (3). Western blots were probed with the antibodies indicated. (C) Scheme for purification of multiple MLL family complexes from HeLa extracts expressing a FLAG (f-) tagged WDR5 protein. (D) GST-PAF1C specifically interacts with components of the MLL1 complex. The f-WDR5 preparation was subjected to GST pulldowns in the presence of either GST or a GST-hPAF1C and the results were blotted and probed with the antibodies indicated. (E) PAF1 interacts with the MLL1 complex in vivo. 293 nuclear extracts were used for IP’s with the antibodies indicated and western blots were probed with the antibodies indicated. (F) Immunofluorescence of PAF1 (red) and MLL1 (green) in 293 cells. PAF1 and MLL1 overlap in only discrete regions, a blowup of a region of overlap is shown as an example (white box). The white bar represents 5μm. See also Figure S3.

Figure 4. The MLL1 CXXC/RD1 domain interacts directly with PAF1 and PHD finger 3 interacts directly with H3K4Me2/3

(A) PAF1C interacts directly with the CXXC containing RD1 fragment of MLL1. The MLL1 CXXC region was divided into two fragments (RD1 and RD2) and expressed as recombinant GST proteins. Purified GST (1), GST-RD1 (2) and GST-RD2 (3) were used in GST pulldown experiments with a recombinant purified PAF1C. Westerns were probed with the antibodies indicated. (B) The PAF1 component of PAF1C interacts directly with MLL RD1. GST-RD1 was used in a GST pulldown experiment with individual FLAG (f-) tagged components of human PAF1C as shown. The upper panel is a gel stained with colloidal blue and the lower panels are western blots of the same pulldowns using the antibodies indicated. (C) Only full length MLL1-RD1 interacts efficiently with PAF1, while MLL2-RD1 does not. A series of deletions of the MLL1 RD1 region were made and subjected to GST pulldowns with purified f-PAF1. Westerns were probed with the antibodies indicated. (D) Top: conservation with MLL2 and the positions of point mutations made in the MLL1 CXXC domain are shown. Bottom: The R1153A point mutation disrupts the direct interaction between purified GST-RD1 and purified f-PAF1 protein. GST pulldowns were done as in (A). (E) PHD fingers 1 to 4 with and without a point mutation in PHD3 (W1594E) were FLAG and HA double tagged and expressed in 293 cells. Nuclear extracts were subjected to peptide pulldowns as indicated and probed with the antibodies shown. (F) Recombinantly expressed and purified dual PHD3/Bromodomain of MLL1 binds specifically to H3 lysine 4 di and trimethyl peptides (top panel). A W1594E point mutation in PHD3 abolishes this interaction (bottom panel). See also Figure S4.

Figure 5. Binding to H3K4Me and PAF1 are both necessary for recruitment of MLL1 across HoxA9 in vivo

(A) Legend for the ChIP experiments in C. Blue line= HA ChIP in cells transfected with an empty vector, Red line= HA ChIP in cells transfected with the CX-P4 constructs indicated and Grey dots = the binding pattern of CX-P4 with no mutation. (B) PAF1 and H3K4Me interactions are important for recruiting MLL1 to HoxA9. Using α-HA ChIP, recruitment of the MLL1 fh-CX-P4 protein to HoxA9 in Mll1−/− MEF cells (no mutant) was compared to individual PHD3 W1594E, K1176A and R1153A point mutations and a triple mutant. ChIP results shown are typical for at least two independent experiments. Error bars represent standard deviation of three separate PCR reactions. (C) An α-HA western blot showing expression of the wild type and mutant constructs from the experiments in B. W,K,R mut = the W1594E, K1176A, R1153A triple mutant. See also Figure S5.

Figure 6. The MLL-AF9 fusion protein needs wild type MLL1, PAF1 and DNA binding for recruitment to HoxA9

(A) Legend for the ChIP experiments in B-J. Blue line= HA ChIP in cells transfected with an empty vector, Red line= HA ChIP in cells transfected with the constructs indicated and Grey dots = the binding pattern of fh-MLL-AF9 in Mll1−/− + MLL cells for comparison. (B) – (J) α-HA ChIP across HoxA9 for the constructs and in the specific cell lines as indicated. ChIP results shown are typical for at least three independent experiments. Error bars represent standard deviation of three separate PCR reactions. See also Figure S6.

Figure 7. Recruitment of fh-MLL-AF9 to HoxA9 in bone marrow cells is essential for leukemic transformation

(A) Leukemogenic transformation of MLL-AF9 constructs. Retrovirally expressed FLAG and HA tagged (fh-) MLL-AF9 fusion protein constructs that were either normal for MLL-AF9 or had R1153A (PAF1 disruption), or K1176A (DNA binding disruption) point mutations were transduced into murine bone marrow cells, along with empty vector, untagged MLL-ENL, and triple FLAG tagged NUP98-Jarid controls. Error bars represent the standard deviation of averaged, cumulative cell numbers across five independent lines for each construct. (B) HoxA9 expression in transformed and control cell lines. Cells from (A) were tested for HoxA9 expression and the average values across all five independent cell lines per construct are shown. Error bars represent standard deviation. HoxA9 expression in Mll1−/− and Mll1−/− + MLL cell lines are also shown for comparison. (C) ChIP in progenitor enriched (Blue line) and fh-MLLAF9 (line #1, Red line) cells with the antibodies indicated. ChIP experiments in Mll1−/− + MLL cells are shown for comparison (Grey dots). (D) PAF1 is required for MLL1 and MLL-AF9 binding to HoxA9. The fh-MLLAF9#1 cell line was treated with control vs PAF1 siRNA’s and the cells were subjected to western blot (inset) and ChIP analysis using the antibodies indicated. α-HA antibody recognizes the HA tagged MLL-AF9 fusion protein. ChIP results were quantified using a primer/probe set for the silent Hoxc8 locus or the mouse HoxA9 coding region (primer/probe set 5 from D). ChIP signal was quantified to inputs and then relative to the HoxA9 control signal which was arbitrarily set to 100. (E) 1= wild type MLL1 is recruited to HoxA9 through interactions with H3K4Me, PAF1 and DNA. 2= MLL1-mediated activation of the HoxA9 locus disrupts binding of repressors such as ESET and creates a more “open” chromatin conformation. 3= An MLL1 fusion protein can now bind through interactions with PAF1 and with CpG rich DNA. ChIP results shown are typical for at least two independent experiments. Error bars represent standard deviation of three separate PCR reactions. See also Figure S7.