Colorectal cancer hot spot mutations attenuate the ASXL-MLL4 interaction

Abstract

Human additional sex combs like (ASXL) proteins are involved in the maintenance of both transcriptional activation and repression through their ability to bind multiple chromatin regulators, including two tumor suppressors: deubiquitinase BAP1 and methyltransferase MLL4 (KMT2D). The ASXL genes are often altered in colorectal cancer (CRC), and ASXL1 is one of the four hub genes related to the pathogenesis of CRC. Here, we show that MLL4 and BAP1 interdependently target specific genomic regions and positively or negatively regulate expression of a subset of genes in the human colon carcinoma HCT116 cells. MLL4 and BAP1 colocalize on a subset of enhancers and promoters in an interdependent manner. Genomic distribution of BAP1 in CRC cells differs from that in ESCs, with substantially more BAP1 binding sites identified on enhancers and promoters in HCT116 cells. MLL4 occupancy on MLL4⁺ BAP1⁺ genomic regions depends on functional ASXLs that interact with both MLL4 and BAP1, and CRC-relevant mutations attenuate the formation of the MLL4-ASXL complex. Mutational analysis and binding assays identified CRC hot spot mutations in ASXLs. Our findings suggest that alterations in the genomic distribution of the MLL4-ASXL-BAP1 axis and CRC hot spot mutations in ASXLs perturb normal transcriptional programs and may trigger pathogenic events in colon cancer.

Regulation of gene transcription depends on a fine-tuned balance of activities of Trithorax group proteins that activate expression and repressive Polycomb group proteins that silence it. The three-member additional sexcombs-like (ASXL1/2/3) family of proteins is a distinctive group of transcriptional coregulators characterized by dual activity. ASXLs are involved in the maintenance of both transcriptional activation and repression through their ability to bind multiple ligands, such as the deubiquitinase BAP1 (BRCA-1 associated protein 1) to form the PR-DUB complex (1, 2, 3). The ASXL genes are frequently found truncated in myeloid malignancies and myelodysplastic syndromes or mutated in solid cancers, including colorectal cancer (CRC) (4, 5, 6). ASXL1 is one of the four hub genes related to the pathogenesis of CRC (7), where ASXL1 was proposed to function as a tumor suppressor (8) or oncogene in a context-dependent manner (5). Loss of ASXL1 has been linked to lymph node metastasis and lymphatic invasion (8). Clinical and pathologic studies show that ASXL1 expression is low in a subset of CRC tissues, and it undergoes frame-shift alterations in 55% of CRC cell lines with microsatellite instability (4, 5, 9). ASXL1 has more somatic mutations in CRC with peritoneal metastasis compared to non-obstructing carcinoma and has emerged as a promising miR-3187-3p target to suppress CRC proliferation (10, 11).

ASXLs share a common domain architecture and along with the DEUBAD domain that associates with BAP1 and stimulates its deubiquitinase activity (2, 12, 13, 14), contain a MLL-binding helix (ASXL_MBH) region—a binding site for the methyltransferase mixed lineage leukemia 4 (MLL4) (15). MLL4, also known as KMT2D, mono-methylates lysine four of histone H3, generating the epigenetic mark H3K4me1 enriched particularly in transcriptional enhancer regions and linked to active gene transcription (16, 17, 18). MLL4 mediates embryonic development and cell differentiation and similarly to ASXLs and BAP1 is recurrently mutated or truncated in various cancers (19). In addition to the catalytic SET domain, MLL4 contains seven plant homeodomain (PHD) fingers with non-redundant functions, enabling this methyltransferase to engage with diverse binding partners and alter the epigenetic landscape of chromatin on demand. For example, the sixth PHD finger of MLL4 recognizes acetylated lysine 16 of histone H4 and the DNA dioxygenase TET3 (20, 21, 22), and the second and third PHD fingers (MLL4_PHD2/3) bind ASXL2_MBH (15), coupling the methyltransferase activity of the tumor suppressor MLL4 to the histone deubiquitinase activity of the tumor suppressor BAP1 (15, 23, 24).

In this study, we report the functional significance of the MLL4-ASXL-BAP1 correlation in human colon carcinoma cell line and identified CRC hot spot mutations that attenuate the binding of ASXLs to MLL4. Our in vivo findings reveal the interdependent genomic co-occupancy of MLL4 and BAP1 in CRC cells, which is essential in the transcriptional regulation of a set of genes.

Results

BAP1 is enriched on H3K4me1⁺ enhancers in HCT116 cells

Although ASXLs have been shown to form complexes with both the deubiquitinase BAP1 and the methyltransferase MLL4(12,15), the importance of ASXLs in genomic distribution of BAP1 and MLL4 in CRC remains unclear. To assess whether ASXLs play a role in regulating MLL4 and BAP1 binding at specific chromatin sites, we generated MLL4-, ASXL2-or BAP1-deficient human colon cancer HCT116 cells using the CRISPR/Cas9 approach. HCT116 cells are near-diploid (25) and therefore highly efficient for gene editing, and additionally, MLL3, a paralog of MLL4, is not expressed in HCT116 cells due to a homozygous frameshift mutation, allowing us to study the effect of MLL4 on the BAP1 genomic distribution without a potential compensatory effect of MLL3.

To determine BAP1 genomic localization, wild-type, and MLL4 knockout (KO) HCT116 cells were infected with lentiviruses expressing doxycycline-inducible, triple T7-tagged BAP1 (BAP1-T7) and then treated with doxycycline for 24 h to induce BAP1-T7 expression. Western blot analysis showed that KO of MLL4 does not affect the expression levels of ectopic BAP1-T7 in HCT116 (Fig. 1A). Unexpectedly, ChIP-seq analysis of BAP1 using a T7 antibody revealed a high degree of occupancy of BAP1 on enhancers, which were defined based on H3K4me1 ChIP-seq signals in wild-type HCT116 cells. Among the 62,590 BAP1 binding regions identified, 28,418 (45%) were located on enhancers and 16,326 (26%) were located on promoters (Fig. 1B). Consistent with previous reports (14, 26), the global levels of H2AK119ub1 were increased by about twofold in BAP1-deficient cells (Fig. 1, C and D). The expression levels of ASXL2 and another PR-DUB complex subunit FOXK1 as well as levels of the enhancer marks H3K4me1 and H3K27ac were mostly unaffected in BAP1 KO cells (Fig. 1, C and D).

Figure 1.

Interdependent genomic binding of MLL4 and BAP1 on enhancers. Wild type (WT) and MLL4 knockout (KO) HCT116 were infected with a doxycycline-inducible lentiviral vector expressing BAP1-T7. Cells were treated with 1 mg/ml doxycycline for 24 h and collected for Western blot using indicated antibodies and ChIP-seq of BAP1-T7 using a T7 antibody. WT, MLL4 KO, and BAP1 KO HCT116 were collected for Western blot and ChIP-seq of MLL4. A, Western blot analyses of nuclear extracts of MLL4 KO cells. B, genomic distribution of BAP1-T7 binding regions. Promoters were defined as transcription start sites ± 1kb. Enhancers were defined as H3K4me1⁺ promoter-distal regions. C and D, Western blots in WT and BAP1 KO HCT116 cells. Whole-cell lysates were analyzed using indicated antibodies (two different anti-BAP1 antibodies, D7W7O and D1W9B, were used). Histone modification levels in (D) were quantified and normalized to H3 of BAP1 KO cells. E, Venn diagram depicting genomic binding of BAP1-T7 and MLL4. F–H, genomic binding of MLL4 and BAP1 on enhancers are interdependent. Heat maps (F) and average profiles (G) for MLL4, BAP1-T7, and H3K4me1 were aligned around the center of MLL4 binding sites on MLL4⁺ BAP1⁺ promoters and enhancers. H, example loci.

Interdependent genomic binding of MLL4 and BAP1 on enhancers

To characterize the genomic distribution of MLL4 and BAP1, we carried out ChIP-seq analysis of MLL4 in wild-type, MLL4-deficient, and BAP1-deficient HCT116 cells. By filtering out non-specific ChIP-seq signals detected in MLL4 KO cells, we identified 16,196 high-confidence MLL4 binding regions (Fig. 1E). Approximately 18% of these regions (2843) overlapped with the BAP1 binding sites and were primarily located on enhancers (2201) (Fig. 1E). Further ChIP-seq analysis in MLL4 KO and BAP1 KO cells revealed a reciprocal genomic colocalization of MLL4 and BAP1 on MLL4⁺ BAP1⁺ promoters and enhancers and on select gene targets (Fig. 1, F–H).

In agreement, KO of ASXL2 resulted in the reduction of MLL4 binding on MLL4⁺ BAP1⁺ promoters and enhancers (Fig. 2). This decrease was not due to the change in expression level of MLL4, as Western blot analysis showed that the depletion of ASXL2 does not affect MLL4 expression (Fig. 2A). In ASXL2-deficient HCT116 cells, the substantial decrease in MLL4 occupancy on enhancers and promoters suggested a role of ASXL2 in the targeting of MLL4 to or stabilization of MLL4 on MLL4⁺ BAP1⁺ regions (Fig. 2, B and C). These results were consistent with the reciprocal ChIP-seq signals of MLL4 and BAP1 observed in BAP1 KO and MLL4 KO cells and indicated that MLL4 and BAP1 contribute to each other’s genomic binding through ASXL2 in HCT116 cells (Fig. 2D).

Figure 2.

Knockout of ASXL2 reduces MLL4 binding on MLL4⁺BAP1⁺enhancers. WT and ASXL2 KO HCT116 were subjected to Western blot and ChIP-seq analyses. A, Western blot analyses. B–D, ASXL2 KO reduces MLL4 binding on MLL4⁺ BAP1⁺ promoters and enhancers. Heat maps (B) and average profiles (B) were aligned around the center of MLL4 binding sites on MLL4⁺ BAP1⁺ promoters and enhancers. D, example loci.

Co-occupancy of BAP1 and MLL4 regulates the expression of a set of genes

Given that MLL4 and BAP1 interdependently regulate their genomic binding on MLL4⁺ BAP1⁺ enhancers, we next asked to what extent this mutual dependence impacts gene expression. For this purpose, we performed RNA-seq in WT and BAP1- or MLL4-deficient HCT116 cells. Of all expressed genes, 677 out of 13,567 were down-regulated >2-fold in BAP1-deficient cells and 436 out of 13,612 were downregulated >2-fold in MLL4-deficient cells (Fig. 3A). We identified 78 genes that show >2-fold decreased expression in both BAP1- and MLL4-deficient cells (Fig. 3B), many of which were functionally associated with RNA metabolism (Fig. 3C). Among these 78 genes, 15 genes were associated with MLL4⁺ BAP1⁺ enhancers (Fig. 3B).

Figure 3.

Genomic colocalization of BAP1 and MLL4 regulates expression of a subset of genes in HCT116. WT, BAP1 KO and MLL4 KO HCT116 were subjected to RNA-seq analyses. A, Pie chart showing gene expression changes in BAP1 KO and MLL4 KO cells. B, genes downregulated >2-fold in BAP1 KO or MLL4 KO cells. C, GO analysis of 78 genes decreased in both BAP1 KO and MLL4 KO. D, genes upregulated >2-fold in BAP1 KO or MLL4 KO cells. E, GO analysis of 136 genes increased in both BAP1 KO and MLL4 KO. F, FPKM values of BAP1 and MLL4 co-regulated genes. G, example loci of BAP1 and MLL4 co-regulated genes. H, Boxplot comparing the mRNA expression levels of MLL4, BAP1, and ASXL1 in normal colon tissues (GTEx, n = 308) versus colon (TCGA-COAD, n = 288). Expression values represent log2-transformed transcripts per million (TPM + 0.001) from RSEM. Statistical significance was assessed using the Mann-Whitney U test. Outliers (log2(TPM + 0.001) < −9) were excluded from visualization.

RNA-seq analysis further revealed that 958 genes were up-regulated >2-fold in BAP1-deficient cells and 713 genes were up-regulated >2-fold in MLL4-deficient cells. Of those, 136 genes showed >2-fold increased expression in both BAP1 KO and MLL4 KO cells, and three genes associated with MLL4⁺ BAP1⁺ active enhancers (Fig. 3D). GO analysis of the up-regulated gene group revealed functional enrichment of these genes in cell differentiation and transcriptional regulation (Fig. 3E). Genomic and transcriptional targets of BAP1 and MLL4 include a set of genes with decreased expression, such as KLK6, LIMA1, ARHGAP29, STEAP1, MGLL, and IVANA, and a set of genes with increased expression, such as CRIP2 and NGFR (Fig. 3, F and G). Collectively, these results indicated both positive and negative regulation of a subset of genes by BAP1 and MLL4 co-bound on MLL4⁺ BAP1⁺ enhancers in HCT116 cells. The fact that MLL4 and BAP1 do not directly interact and are linked through their concomitant binding to ASXLs suggested a possible role of ASXLs in this regulation.

Analysis of expression profiles for MLL4 and BAP1 in human colon cancer (primary tumor) using the datasets from TCGA, TARGET, and GTEx showed that both MLL4 and BAP1 mRNA levels were significantly lower in tumor tissues compared to normal colon tissues, and concurrently ASXL1 was also downregulated (Fig. 3H). We noticed an increased heterogeneity in MLL4 and BAP1 expression across tumor samples. Together these data were in agreement with MLL4 and BAP1 roles as tumor suppressors and suggested potential alterations in normal MLL4/BAP1-dependent transcriptional programs in the context of colon cancer.

CRC-relevant mutations in ASXLs affect the binding of MLL4

ASXL1 is frequently altered in CRC(5,7), amplified in 10% of cases, and mutated in 5% of cases (cBioPortal), and ASXL2 and ASXL3 are altered primarily through mutations (Cosmic). CRC-relevant point mutations A611T, R612M, and A615T were reported in ASXL1, as well as R655I in ASXL2, and R1052M, A1055T, and R1060W in ASXL3 (Cosmic). Several CRC mutations were identified in MLL4, including C276Y, C294R, and V231G (Cosmic). Mapping the CRC-relevant mutations on the structure of MLL4_PHD2/3 in complex with ASXL2_MBH showed that many of these mutated residues are located within the MLL4-ASXL binding interface, prompting us to investigate whether the CRC-relevant mutations impact the interaction between MLL4 and ASXLs (Fig. 4, A and B).

Figure 4.

CRC-relevant mutations are located in the MLL4-ASXLs binding interface.A, a ribbon diagram of the structure of MLL4_PHD2/3 (green and blue) in complex with ASXL2_MBH (orange). Residues mutated in CRC are labeled. B, MLL4_PHD2/3 is shown as surface colored according to electrostatic surface potential ranging from positive; blue (+100 kT/e) to negative; red (−100 kT/e) and generated with PyMol vacuum electrostatics. Bound ASXL2_MBH is shown as a ribbon. Residues mutated in CRC are labeled. C and D, superimposed ¹H,¹⁵N HSQC spectra of MLL4_PHD2/3 collected upon titration with indicated wild type and mutated ASXL1, ASXL2, and ASXL3 peptides. Spectra are color-coded according to the protein: peptide molar ratio.

We produced ¹⁵N-labeled MLL4_PHD2/3 and collected its ¹H,¹⁵N heteronuclear single quantum coherence (HSQC) spectra while ASXL1 peptides (residues 607–627 of ASXL1) wild type or containing the single point mutation A611T, R612M or A615T were added stepwise. Titration of either ASXL1 peptide into the MLL4_PHD2/3 NMR sample led to substantial chemical shift perturbations (CSPs), indicative of the formation of complexes (Fig. 4C). While CSPs induced in MLL4_PHD2/3 by wild-type ASXL1 were in the slow exchange regime on the NMR timescale, CSPs caused by R612M and A615T mutants were in the slow-to-intermediate exchange regime, indicating a decrease in binding activity. Similarly, the slow-to-intermediate exchange regime observed for the association of R655I mutant of ASXL2 (residues 650–670 of ASXL2) or R1052M mutant of ASXL3 (residues 1047–1067 of ASXL3) with MLL4_PHD2/3 suggested a reduced binding (Fig. 4D).

The binding capabilities of the CRC-relevant ASXLs mutants were measured by microscale thermophoresis (MST) (Fig. 5, A and B). We labeled His-tagged MLL4_PHD2/3 with the RED-tris-NTA fluorophore and monitored the fluorescence intensity using a direct binding assay in which increasing amounts of ASXL peptides were added to each MLL4_PHD2/3 sample. In support of NMR data, R612M in ASXL1 and R1052M in ASXL3 reduced binding of MLL4_PHD2/3 ∼4- and ∼2.5-fold, respectively, and we previously reported that the R655I mutation in ASXL2_MBH decreases this interaction ∼16-fold (15). Additionally, the association of MLL4_PHD2/3 with the A615T mutant of ASXL1 was reduced ∼3-fold. Reciprocally, binding of the V231G mutant of MLL4_PHD2/3 to wild-type ASXL2 was also reduced ∼3-fold (Fig. 5B). Two other CRC-relevant mutants of MLL4_PHD2/3 C276Y and C294R were not tested as mutations of these zinc-coordinating cysteine residues likely disrupt the MLL4_PHD2/3 structure and therefore abolish binding to ASXLs. Together, the binding data suggested that CRC-relevant mutations located at the ASXL_MBH -MLL4_PHD2/3 binding interface have a negative effect on the complex formation.

Figure 5.

CRC-relevant hot spot mutations in ASXLs.A and B, MST binding curves (A) used to determine binding affinities (B) of wild type and mutated MLL4_PHD2/3 for the indicated ASXL1, ASXL3 and ASXL2 peptides. The K_d values represent average of four independent measurements, and errors represent standard deviation. n = 4 Point errors in (A) represent SEM. (^a) taken from (15). C, alignment of amino acid sequences of ASXLs. CRC-relevant hot spot mutations are marked with red boxes. D and E, overlayed ¹H,¹⁵N HSQC spectra of MLL4_PHD2/3 in the absence (black) and presence of a 5-fold molar excess of indicated mutated ASXLs peptides. F, a model for the ASXL-dependent colocalization of the H3K4me1-specific methyltransferase MLL4 and the H2AK119ub-specific deubiquitinase BAP1 at enhancer and promoter regions of a subset of target genes.

Hot spot CRC mutations

Alignment of the sequences of ASXL1, ASXL2, and ASXL3 shows that R612 of ASXL1 is aligned with R655 of ASXL2 and R1052 of ASXL3, whereas A615T of ASXL1 is aligned with A1055T in ASXL3 (Fig. 5C). Further analysis of CSPs in NMR titration experiments indicated that all wild-type and mutated ASXL peptides occupied the same binding pocket of MLL4_PHD2/3, as essentially the same set of resonances of the apo-state of MLL4_PHD2/3 was perturbed most (Fig. 5, D and E). The finding that R612/R655/R1052 is recurrently mutated in all three ASXLs to a hydrophobic residue suggested that this site may represent a hot spot mutation in CRC. The side chain guanidino group of this arginine (R655) in the ASXL2-MLL4_PHD2/3 complex is restrained through hydrogen bonds and salt bridges with the carboxyl groups of E292 and D257 of MLL4_PHD2/3, therefore loss of these charge-charge contacts by mutating arginine to a hydrophobic methionine or isoleucine can explain the decrease in binding (Fig. 4A). The carboxyl group of D257 also restrains the side chain of K663 of ASXL2 (R1060 in ASXL3), hence a ∼2-fold decrease in binding activity of the R1060W mutant of ASXL3 can be attributed to lacking the stabilizing contact with D257 and/or steric hindrance. Likewise, the recurrent mutation A615T in ASXL1 and A1055T in ASXL3 may be another CRC hot spot mutation. The corresponding residue in ASXL2, A658, is bound in a small hydrophobic pocket, and adding even a small hydroxyl group of threonine may lead to steric hindrance and a decrease in binding affinity (Fig. 4B).

Discussion

Colorectal carcinoma is the third most commonly diagnosed cancer, the second leading cause of mortality among cancer patients, and the most lethal malignancy in men younger than 50 years (27). While in the past few decades the overall CRC mortality slowly declines due to widespread preventive screening and advances in pharmacological treatments, the CRC burden is alarmingly shifting to the younger population with the advanced stage of disease (27). Among several etiologic mechanisms underlying CRC development, changes in human epigenome and alterations in gene expression programs have emerged as one of the major contributors to tumor initiation and progression. In this work, we show that the key components of the epigenetic machinery, the histone methyltransferase MLL4 and the histone deubiquitinase BAP1 interdependently target enhancer and promoter regions and regulate expression of a subset of genes in the human colon carcinoma cell line HCT116. MLL4 and BAP1 colocalize to enhancers and promoters in a reciprocal manner, corroborating the prevalent role of MLL4 in binding to active enhancers observed in mouse embryonic stem cells (ESCs) (15) (Fig. 5F). However, genomic distribution of BAP1 in HCT116 differs substantially from that of ESCs, with more BAP1 binding sites being identified on enhancers and promoters, i.e. 45% on enhancers and 26% on promoters in HCT116 cells, comparing to 38% on enhancers and 16% on promoters in ESCs.

The MLL4 and BAP1 co-occupancy depends on ASXL proteins that directly interact with both MLL4 and BAP1 and bring together two enzymatic activities to specific genomic loci. ASXLs have been shown to contribute to the pathogenesis of CRC(5,7,8,19), and our studies demonstrate that the recurrent CRC hot spot mutations in ASXLs attenuate the formation of the complex with MLL4. These include mutation of arginine to a hydrophobic residue, as well as mutation of alanine to threonine in the amino-terminal part of the α-helix in ASXL_MBH. In this respect, given the clinical importance of ASXLs in CRC, genome editing approaches to correct these mutations and restore the expression of wild-type ASXLs might be beneficial and necessitate thorough examination. Further studies are also needed to distinguish the biological roles of each ASXL homolog and define the potential functional redundancies within this family. ASXLs act as scaffolding proteins linking not only MLL4 and BAP1 but also the demethylase LSD1 and various nuclear hormone receptors and are involved in the recruitment of the Polycomb-group repressor complex (PRC) and the trithorax-group activator complex (1, 3, 6, 28). Likewise, MLL4 has other ligands, including the DNA dioxygenase TET3 (22). To better understand the multifaceted roles of the MLL4-ASXL-BAP1 axis in transcriptional regulation, forthcoming studies should also be focused on the potential additional effects that are not accounted for by the direct interaction of ASXLs and MLL4, such as mutations in other regions of these genes (Cosmic, cBioPortal, DepMap), establishing a functional crosstalk involving other binding partners of ASXLs, MLL4 and BAP1, and investigating how aberrant activities of these major epigenetic proteins and their complexes are linked to other human diseases.

Experimental procedures

Generation of ASXL2, BAP1, or MLL4 knockout HCT116 cell lines

HCT116 human colon cancer cells were cultured in McCoy’s 5A medium supplemented with 10% FBS. Knockout cell lines were generated using the CRISPR/Cas9 system. The sgRNA sequences targeting ASXL2 (5′-ACTTTACCTGCTGGAATGG-3′) or MLL4 (5′-TTCTCAAACCACTCCGAGT-3′) were cloned into lentiCRISPR v2 plasmid (Addgene #52961). The sgRNA targeting BAP1 was obtained from Addgene (#125837). To generate knockout cells, HCT116 was transfected with lentiCRISPR v2 containing sgRNA sequences using Lipofectamine 3000 (Thermo Fisher). 24 h later, cells were selected with 1.5 μg/ml puromycin for 2 days and recovered for 2 days in culture media without puromycin. 1 × 10⁴ cells were plated into a 15 cm dish. 1 week later, colonies were picked and transferred into a 96-well plate. Knockout clones were verified by Sanger sequencing.

Generation of cell lines expressing BAP1-T7

Wild type or MLL4 knockout HCT116 cell lines were infected with Doxycycline-inducible lentiviral pCW57.1 vector (Addgene #41393) expressing triple T7-tagged BAP1. 24 h later, cells were selected with 1.5 μg/ml puromycin for 4 days and recovered for 2 days in culture media without puromycin. Cells were treated with 1 μg/ml Doxycycline for 24 h and collected for Western blot and ChIP-seq analyses.

Western blot

For extracting nuclear proteins, cells were washed once with ice-cold PBS, collected in buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT, and 0.2 mM PMSF), resuspended in buffer B (buffer A + 0.1% NP40, 0.5 mM DTT and 0.2 mM PMSF) and incubated on ice for 10 min. After centrifugation at 1000g, resulting nuclei were resuspended in buffer C (20 mM HEPES, pH 7.9, 1.5 mM MgCl₂, 420 mM NaCl, 0.2 mM EDTA, and 25% glycerol, 0.5 mM DTT and 0.2 mM PMSF). For extracting histones, cells were washed once with ice-cold PBS, collected in hypotonic buffer (10 mM Tris-HCl pH 8.0, 1 mM KCl, 1.5 mM MgCl₂, 0.5 mM DTT, and 0.2 mM PMSF) and incubated on ice for 10 min. After centrifugation at 1000g, histones were extracted with 0.2 N HCl. For extracting whole cell lysates, cells were collected and lysed in RIPA buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.1% SDS, 1% NP 40, 0.5% sodium deoxycholate) supplemented with protease inhibitors for 20 min, and subjected to sonication. Extracted proteins were resolved using 4 to 15% Tris-Glycine gradient gels (Bio-Rad Laboratories) or 3 to 8% Tris-Acetate gradient gels (Invitrogen). Primary antibodies used in Western blot were indicated as follows. Anti-MLL4 and anti-UTX were home-made. Anti-T7-Tag (#13246), anti-BAP1 (D7W7O, #13271), anti-BAP1 (D1W9B, #13187) and, anti-H2AK119ub1 (#8240) were from Cell Signaling Technology. Anti-RbBP5 (A300-109A) and anti-ASXL2 (A302-037A) were from Bethyl Laboratories. Anti-FOXK1 (ab18196), anti-H3 (ab1791), and anti-H3K27ac (ab4729) were from Abcam. Anti-BRG1 (sc-10768 X) was from Santa Cruz Biotechnology. Anti-H3K4me1 (13–0040) was from EpiCypher.

ChIP-seq library preparation

Cells were cross-linked with 1% formaldehyde for 10 min. Cross-linking reaction was stopped by 125 mM glycine for 10 min 30 to 40 million cross-linked cells were resuspended in 5 ml lysis buffer (5 mM PIPES, pH 7.5, 85 mM KCl, 1% NP-40 and protease inhibitors), incubated on ice for 10 min, and spun down at 1000g for 5 min at 4 °C. The pellet was washed once with 5 ml lysis buffer. Resulting nuclei were resuspended with 2 ml TE buffer (50 mM Tris-HCl, pH8.0, 10 mM EDTA and protease inhibitors) and sonicated for 10 min (30 s on and 30 s off) at 20% amplitude. After adding detergents to a final concentration of 1% Triton-X100, 0.1% SDS, 0.1% sodium deoxycholate (1× RIPA), sonicated chromatin was spun down at 15,000g for 10 min at 4 °C. For each ChIP, 10 μg of target antibodies and 2 μg of spike-in antibody (anti-H2Av, Active Motif, #61686) were mixed with 400 to 600 mg sonicated chromatin and incubated on a rotator at 4 °C overnight. Next day, 50 μl prewashed protein A Dynabeads (Thermo Fisher) were added to chromatin-antibody complex and incubated for 2 h at 4 °C. ChIPed samples were washed twice with 1 ml RIPA, twice with 1 ml RIPA + 300 mM NaCl, twice with 1 ml LiCl buffer and once with PBS. Samples were eluted in 100 μl buffer containing 0.1 M NaHCO₃, 1% SDS, and 20 μg Proteinase K at 65 °C overnight and DNA was purified using QIAquick PCR purification kit (Qiagen). Total ChIPed DNA or 1 μg of input DNA were used for ChIP-seq library construction using the NEBNext Ultra II DNA Library Prep kit with dual index primers and AMPure XP magnetic beads (Beckman Coulter). Library quantity was estimated with Qubit assays. The final libraries were sequenced on Illumina NovaSeq 6000.

RNA-seq library preparation

Total RNA was extracted using TRIzol (Life Technologies). 1 μg total RNA was used to purify mRNA using the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB, E7490). Isolated mRNAs were reverse-transcribed into double stranded cDNAs and subjected to RNA-seq library construction using the NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB, E7770) according to the manufacturer’s instructions. Final libraries were sequenced on Illumina NovaSeq 6000.

Computational analysis

ChIP-seq peak calling

Raw sequencing data were aligned to the human genome hg38 and the drosophila genome dm6 using Bowtie2 (v2.3.4) (29). To identify ChIP-enriched regions, SICER (v2) was used (https://zanglab.github.io/SICER2/). For ChIP-seq of MLL4 and BAP1-T7, the window size of 50 bp, the gap size of 50 bp, and the FDR threshold of 10⁻³ were used. For ChIP-seq of H3K4me1, the window size of 200 bp, the gap size of 200 bp, and the false discovery rate (FDR) threshold of 10⁻³ were used.

Regulatory regions

Promoter was defined as transcription start sites ±1 kb. Promoter-distal regions were divided into enhancer (H3K4me1⁺) or other (H3K4me1⁻).

Heat maps

The heat map matrices were generated using in-house scripts with 50 bp resolution and visualized in R using the gplots package. Genomic regions shown in the heat maps were ranked according to the intensity of MLL4 at the center of the 400 bp window in WT cells. Average profiles were plotted using the normalized ChIP-seq reads in 5 bp bins from the center of the MLL4 binding site on enhancers or promoters to 2 kb on both sides.

RNA-seq data analysis

Raw sequencing data were aligned to the human genome hg38 using a STAR aligner (30). Reads on exons were collected to calculate fragments per kilobase per million (FPKM) as a measure of gene expression level. Only genes with exonic reads of FPKM >1 were considered expressed. For comparing gene expression levels in WT and MLL4-or BAP1-deficient cells, average FPKM values from duplicates were calculated, and fold change cutoff of >2 was used to identify differentially expressed genes. Gene ontology (GO) analysis was done using DAVID (https://david.ncifcrf.gov).

mRNA expression in colon adenocarcinoma

Gene expression data for normal colon tissues (GTEx) and colon adenocarcinoma (TCGA-COAD) were retrieved from the UCSC Xena Browser (https://xenabrowser.net/datapages/). Cell line data and non-colon tissues were excluded. RNA-seq values are reported as log2-transformed transcripts per million (TPM + 0.001) using RSEM normalization. To mitigate artifacts, outliers with log2(TPM + 0.001) values below −9 were filtered prior to analysis. Differential expressions between normal and tumor groups were assessed using the non-parametric Mann-Whitney U test.

Protein expression and purification

The PHD2 and PHD3 fingers of human MLL4 (aa 227–324) were cloned into a pGEX-6p-1 vector. Mutant construct MLL4_PHD23 V231G and His-tag insertion (C-terminal 6xHis tag) MLL4_PHD23 were generated using the QuikChange Lightning kit (Stratagene). All constructs were confirmed by DNA sequencing. All constructs were expressed in Rosetta2 (DE3) pLysS competent cells. Protein production was induced with 0.5 mM IPTG overnight at 16 °C in Luria broth (LB) or minimal media (M9) supplemented with ¹⁵NH₄Cl and 50 μM ZnCl₂. The GST-tagged proteins were purified on Pierce Glutathione Agarose beads (ThermoScientific) in 50 mM Tris (pH 7.5) buffer, supplemented with 150 mM NaCl and 5 mM dithiothreitol (DTT). The GST tag was cleaved with PreScission protease overnight at 4 °C. All unlabeled proteins were further purified by size exclusion chromatography (SEC) and concentrated in Amicon centrifugal filter units (MilliporeSigma).

NMR

NMR experiments were performed at 298 K on a Bruker 600 MHz spectrometer equipped with a cryogenic probe. Experiments were carried out using 0.1 mM uniformly ¹⁵N-labeled proteins in 50 mM Tris, pH 6.8 buffer, supplemented with 150 mM NaCl, 2.5 mM DTT, and 8% D₂O. ¹H,¹⁵N heteronuclear single quantum coherence (HSQC) spectra were recorded in the absence and presence of increasing concentrations of peptides.

MST

Microscale thermophoresis (MST) experiments were carried out on a Monolith NT.115 instrument (NanoTemper). All experiments were performed using SEC purified MLL4_PHD23-6xHis protein in a 25 mM Tris-HCl pH 7.0 buffer, containing 150 mM NaCl and 5 mM DTT. MLL4_PHD23-6xHis was labeled using a His-Tag Labeling Kit RED-tris-NTA (second Generation, NanoTemper) and kept constant at 10 nM. Dissociation constants were determined using a direct binding assay in which peptide was varied in concentration by serial dilution of discrete samples. The measurements were performed at 40% LED and medium MST power with 3 s pre-laser time, 20 s laser on time, and 1 s off-time. The K_d values were calculated using MO Affinity Analysis software (NanoTemper) (n = 4). Plots were generated in GraphPad PRISM.

Data availability

ChIP-seq and RNA-seq datasets described in this paper have been deposited in NCBI Gene Expression Omnibus under accession number GSE213564. All other relevant data supporting the key findings of this study are available within the article.

Code availability

This paper does not report original code.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Brian D. Strahl