DISORDERED HISTONE METHYLATION IN HEMATOLOGICAL MALIGNANCIES THE CASE OF UTX/KDM6A

Abstract

Alterations of epigenetic proteins that modulate the gene repressive lysine 27 on histone H3 (H3K27me) are recurrent features in cancers, including multiple myeloma (MM). The histone demethylase UTX/KDM6A, mutated in up to 10% of cases of MM activates genes by removing the H3K27me3 repressive histone mark, counteracting EZH2. RNA-sequencing studies showed that UTX upregulated genes in association with loss of H3K27me. Treatment of MM cell lines with an EZH2 inhibitor preferentially slowed growth of UTX-null cells. EZH2 inhibitors activated many of the same genes as UTX but also induced the earlier stage B cell marker Bcl6 which, in turn, shut off the late B cell IRF4 and MYC, leading to cell death.

INTRODUCTION

In eukaryotic cells, DNA is complexed with histone proteins to form the nucleosome, consisting of ~145 bp of DNA wrapped around an octamer of the four core histone proteins: H2A, H2B, H3, and H4 (1).The histone octamer consists of a central (H3-H4)₂ tetramer flanked on either side by two H2A-H2B dimers (2). Nucleosomes have a general inhibitory effect on transcription of RNA into DNA and are important for maintaining the structural integrity of DNA as they protect DNA from damage during interphase (3,4). Nucleosomes are the basic repeating structural elements that allow folding of chromosomal DNA into the higher order structure of chromatin, and they maintain genome structure during condensation of chromatin into chromosomes and segregation of chromatids during cell division (5,6). Histone octamers form nucleosomes across the genome in a nonrandom pattern according to a “nucleosome code” and tend not to form on AT-rich DNA which is more difficult to bend (7). This code and nucleosome positioning allows for the occlusion of critical DNA binding sites for transcription factors and can inhibit the formation of the pre-initiation complex. In the context of chromatin, transcription is a complex process achieved by multiprotein machines that 1) recognize enhancer and transcriptional start sites, 2) recruit chromatin modifying cofactors to these sites to, 3) modify histone and DNA components changing chromatin physicochemical properties and creating binding sites for other transcriptional machines. 4) The mediator complex and other cofactors bridge enhancer and promoter regions, and 5) the cohesin complex allows loops to form between promoters and enhancers. 6) Short bidirectional enhancer-derived transcripts assist in the loading of cohesin onto these nascent loops. The formation of a productive loop and enhancer-promoter interaction 7) poises RNA polymerase II for transcription. 8) Adenosine triphosphate–dependent chromatin remodeling complexes physically move nucleosomes and distort chromatin and 9) DNA is unwound to allow RNA polymerase to make a RNA template of DNA. 10) RNA polymerase often “stalls” a short distance 3’ of the start site of transcription, but enhancer-promoter interactions encourage the recruitment of 11) elongation factors that drive transcription.

Changes in gene expression are accompanied by changes in chromatin. Activated genes have enhancers bound by p300/CBP HATs and display acetylation of histone tail residue 27 (H3K27Ac) and monomethylation of H3K4 (H3K4me1). Active promoters show H3K4me3, H3K36me2, acetylation of H3K9, H3K27, and H3K36me3 in transcribed gene bodies. Repressed genes lose H3K27Ac at enhancers, lose H3K4me3 at promoters, and display H3K27me3 and H3K9me3 (Reviewed Segal [8]). Disruption of the epigenetic patterns by mutations of epigenetic regulators is considered to a driver of malignancy and restoring or exploiting such epigenetic changes may represent a potential therapeutic modality in cancer. Many of the changes in epigenetic chromatin modification affect H3K27 (Reviewed in Nichol et al. [9]). For example, our group and collaborators showed that EZH2 gain of function mutations in lymphoma cause a global increase in H3K27me3 levels and repression of genes required for B cell differentiation (10). In multiple myeloma (MM), 15% of cases harbor an immunoglobulin gene translocation (chromosomal translocation t(4;14)) to the NSD2/MMSET gene encoding a histone methyltransferase, which causes a genome-wide increase in H3K36 dimethylation and decrease in the mark H3K27me3. The result is a loss of gene repression and activation of an aberrant gene program that stimulates cell growth. In ~10% of cases of childhood acute lymphocytic leukemia, gain of function mutations that increase the activity of the enzyme and yield a similar result to t(4;14) with an increase in H3K36me2 at the expense of H3K27me3 (11,12). These changes are predicted to cause activation of genes by precluding the repressive action of the H3K27me3 mark.

KDM6A/UTX is a histone demethylase that can remove the repressive H3K27me3 modification (13). KDM6A is part of a multiprotein complex that includes the histone acetyl transferases p300 or CBP, the enhancer active histone methyltransferase KMT2D or KMT2C and SWI/SNF chromatin remodeling proteins with adenosine triphosphatase activity (Figure 1). Encoded on the X chromosome, KDM6A is a pseudoautosomal gene suggesting that its gene dosage is critical for normal cell and organismal development. KDM6A isoform is 1400 aa long and contains tetratricopeptide repeats involved in protein-protein interactions (14) and the JmJC demethylase domain which removes methyl groups from histone using Fe²⁺ and alpha-ketoglutarate as a cofactor. KDM6A and partners are required for proper activation of enhancers although the relative role of the demethylase activity versus the scaffolding properties of the protein is uncertain. In animal development, some phenotypes resulting from KDM6A knockout can be restored by repletion with a demethylase inactive form of the enzyme (15,16). Additionally, the Y chromosome encoded homologue of KDM6A, UTY has very weak demethylase activity in vitro and in vivo (17). Other evidence that KDM6A has a large role as an enhancer active scaffolding protein comes from human genetics. Heterozygous germline mutations of the KMT2D histone methyltransferase cause the Kabuki syndrome, a developmental disorder characterized by facial and growth anomalies and intellectual impairment. Five percent of patients with this syndrome have KDM6A not KMT2D mutations (18), indicating that KDM6A loss phenocopies KMT2D loss.

Fig. 1.

UTX/KDM6A is present in a complex that contains the histone acetyltransferase(s) CBP/p300, the chromatin remodeling SWI/SNF proteins and the KMT2C/D. Together they activate transcription by removing the repressive H3K27me3 mark and adding the H3K4me1 and H3K27Ac mark at enhancers.

Inactivating mutations of KDM6A as well as KMT2C, KMT2D, and SWI/SNF components of its complex are among the most common found in all human cancers (19,20). UTX lesions tend to be homozygous in females and to be accompanied by the loss of its paralog UTY in males, suggesting a tumor suppressor role (21). Supporting this idea, loss of UTX promotes proliferation in many contexts, and accelerates NOTCH1-driven T-ALL onset in vivo (21–24). Nevertheless, the role of UTX in cancer seems to be tissue-specific as overexpression of UTX in breast cancer promotes proliferation and invasion (25). In agreement with this, UTX target genes seem to be very different among cell types, suggesting a cell-specific role (22,24,25). In MM, disruptive mutations encompassing the KDM6A locus are found in 3% to 4% of primary MM (21,26) and were associated with decreased survival (21,26,27). Copy number variation analyses indicate loss of this gene in approximately 25% MM cases (https://research.themmrf.org). Of note, ~40% of MM cell lines, which tend to be derived from patients with advanced disease and extramedullary spread, are UTX deficient.

Here I discuss recent work from my group showing the importance of UTX and deregulating chromatin control of gene expression. Mutations of this epigenetic enzyme appear to be a clear driver of the malignant phenotype. In the case of UTX, this loss of function mutation can be potentially alleviated by a synthetic lethal approach targeting another chromatin pathway that counteracts the H3K27me3 demethylase, namely, the EZH2 H3K27me3-specific histone methyl transferase.

MATERIALS AND METHODS

Cell Culture

The isogenic MM cell line pairs ARP-1 and ARD were cultured in advanced Roswell Park Memorial Institute medium supplemented with 4% fetal bovine solution and GlutaMAX (Thermo Fisher Scientific, Waltham, Massachusetts). Activator of RNA decay (ARD) cells were transduced with pInducer20 harboring a Tet operator-controlled UTX cDNA (pUTX). Lentiviruses were generated by transfection of 293T cells with this plasmid and vectors psPAX2 and pMD2.G (Addgene) (28), using Fugene 6 (Roche Applied Science, Indianapolis, Indiana). Lentiviruses incubated with cells along with of 6 µg/ml polybrene (Millipore, Darmstadt, Germany) followed by selection in G418 (Corning). For the induction of UTX, cells were grown in presence of 25 ng/mL of doxycycline. For the proliferation assays, 100,000 cells per well were seeded in 2 ml of media in 6-well plates +/− doxycycline. Every 3 days, live cells were counted by the trypan blue exclusion method, and the initial number of cells was replated in fresh media with or without doxycycline.

RNA Sequencing

RNA was extracted from ARP-1 and ARD cells; the add-back system treated with doxycycline for 3, 6, and 9 days; and ARP-1 and ARD cells were treated with 4 µmol/l of GSK343 or GSK669 for 7 days. Total RNA was sequenced on HiSeq 2000 (Illumina, San Diego, CA). The Bioconductor package, edgeR (v3.8.5) was used to identify differentially expressed genes. Genes with less than one read per million of sequenced reads in three or more samples were filtered out. The Benjamini and Hochberg false discovery rate was set at <0.05, and fold change > 1.5. Ingenuity Pathway Analysis software (Qiagen, Redwood City, CA) identified biological pathways.

Mouse Models

Animal experiments were approved in compliance with the Northwestern University Animal Care and Use Committee. Six-week-old female C57BL6 Nu/Nu mice (Jackson Laboratory, Bar Harbor, ME) were injected with 5 × 10⁶ cells transduced with a plasmid harboring the luciferase gene (pFU-2LT) in 100 µl cold phosphate buffered saline, mixed with 100 µl of CultreX PathClear BME (3432-005-02, Trevigen, Gaithersburg, MD) subcutaneously. When tumors were detectable, half of the mice (n = 7) were administered doxycycline 2 mg/ml. Luciferin (150 mg/kg) (Gold Biotechnology, St. Louis, MO) was injected intraperitoneally and images were obtained using IVISR Spectrum (Caliper Life Sciences, Inc., Hopkinton, MA) and quantified using Living Images software. As indicated, mice were treated with daily intraperitoneal injections of GSK126 50 mg/kg (Activebiochem), or vehicle (20% captisol, pH 4.5). Tumor growth was measured weekly as above.

RESULTS

ARD- and a UTX-deficient cell line was transduced with a tetracycline inducible form of UTX. Cells were replated every 3 days and population doublings determined by the formula 3.32 log (viable cells at harvest/viable cells at seed) and plotted over time (Figure 2). Re-expression of UTX led to fewer population doublings and a decreased cell accumulation over time. When these cells tagged with luciferase were inoculated into immunosuppressed mice, growth suppression was more pronounced with decreased tumor growth as measured by bioluminescence, and decreased cell division was evidenced by a marked reduction in staining with the cell cycle marker Ki67 (Figure 3).

Fig. 2.

Effect of UTX on growth of myeloma cells in vitro. ARD cells conditionally expressing UTX were cultured over a 12-day period diluting cells every 3 days and courting the viable cell yield by trypan blue exclusion. Adapted from Ezponda et al. (32).

Fig. 3.

Effect of UTX on growth of myeloma cells in vivo. ARD cells conditionally expressing UTX were injected into Nude mice and given doxycycline to induce UTX expression. Tumor growth was measured by bioluminescence. Cell proliferation in tumors was quantified by Ki67 staining. Adapted from Ezponda et al. (32).

To determine the molecular basis of changes in growth as well as other changes noted such as decreased cell adhesion and migration upon re-expression of UTX, RNA-sequencing analysis was performed on ARD cells before and after the induction of UTX. These results showed that a panel of genes was largely downregulated when comparing UTX replete cells. Re-expression of UTX led to regulation of a panel of genes involved in adhesion and signaling (Figure 4). Loss of UTX was associated with a gene expression signature akin to the epithelial mesenchymal transition including loss of expression of E-cadherin and increased expression of alternative adhesion genes.

Fig. 4.

Gene expression modulated by UTX. (A) Heat map indicating genes expressed in UTX replete ARP1 cells, downregulated in ARD, UTX deficient cells and re-expressed with add-back of UTX. (B) Gene pathways regulated by UTX. Adapted from Ezponda et al. (32).

We hypothesized that the loss of UTX would, at certain genes, lead to a predominant action of the repressive polycomp complex 2 and its core enzyme EZH2. Loss of UTX then could be overcome by inhibiting EZH2 action. Over the past several years a number of EZH2 inhibitors have been developed (29–31). We used one such inhibitor GSK 343, applying the compound to ARD cells and measuring changes in gene expression by RNA sequencing. We found that EZH2 inhibition activated a number of genes also turned on by re-expression of UTX (Figure 5). These include genes involved in adhesion as well as an additional set of genes involved in cell life death decisions. Whereas UTX re-expression modestly slowed cell growth, EZH1 inhibition led to decreased viability of UTX null cells. This was associated with increased expression of the early B cell marker Bcl6 and repression of the late B cell gene IRF4 and the critical cell proliferation gene c-Myc (32). If UTX is a tumor suppressor and it is lost in myeloma and other tumors, a rational way to rebalance gene expression and restore normal cell growth would be to inhibit EZH2. We tested this idea in a pre-clinical experiment in which mice were injected in the wither flank with UTX wild-type (wt) and UTX mutant myeloma cells tagged with luciferase. The mice were treated with vehicle or GSK126, an inhibitor compound similar to GSK 339. In accordance with gene expression data, EZH2 inhibition mimicked the repletion of UTX and substantially decreased growth of the UTX null myeloma cells in vivo (Figure 6).

Fig. 5.

Gene expression in UTX null multiple myeloma cells modulated by EZH2 inhibition. (A) Heat map indicating genes activated and repressed in response to EZH2 inhibitor in ARD UTX null cells. (B) Overlap of genes regulated by UTX and EZH2 inhibition. Adapted from Ezponda et al. (32).

Fig. 6.

EZH2 inhibition blocks UTX null myeloma growth in vivo. (A) Schedule of treatment of mice injected with UTX wild-type ARP1 cells on the left flank and UTX null ARD cells in the right flank. (B) Bioluminescence measurement of tumor volume (top). Measurement of tumor volumes in the presence or absence of EZH2 inhibitor treatment (bottom). Adapted from Ezponda et al. (32).

DISCUSSION

Loss of UTX occurs in many cancers (21,33,34), and is common in advanced bladder cancer and high-risk acute lymphoid leukemia (33,35,36), suggesting that it could drive the biological behavior of such malignancies. UTX disruption is not common in newly diagnosed MM but is common in MM cell lines, which are frequently derived from advanced MM which has spread beyond the bone marrow. This background information and our data suggest that UTX loss may contribute to the progression and dissemination of MM. Loss of UTX leads a gene expression pattern similar to the epithelial-to-mesenchymal transition, which in a mouse myeloma xenograft model was associated with tumor progression (37). UTX loss increased expression of adhesion factors such as NCAM1, AOC3, or CDHR5 (data not shown), which may be involved in myeloma dissemination. The increased biological activity of UTX re-expression in vivo relative to in vitro experiments (32) may be due to the effect of UTX on MM stroma/microenvironment interactions.

In contrast with our findings, breast UTX overexpression was associated with increased biological activity of breast cancer (25), suggesting that its role in cancer development may depend on tissue type. Corroborating this notion, UTX target genes identified in our experiments in MM were completely different from those found in experiments in human fibroblasts and breast cancer (24,25). This may be due to the ability to interact with a variety of tissue specific transcription factors (17,38) or operate in an environment where differ subsets of enhancer are set up for gene activation due to such tissue-specific gene regulators.

The loss of UTX in MM cells causes downregulation of a large set of genes whose expression can be rescued by UTX re-expression. This may be due to the failure of the activator complex of UTX, KMT2C/D, and p300/CBP to engage enhancer regions and cannot oppose the repressive action of EZH2 and the PRC2 complex. In support of this notion, we found that nearly half of the genes activated by UTX could also be activated by treatment of UTX null cells with EZH2 inhibitors. Similarly, EZH2 inhibitors change the balance of gene expression in pediatric rhabdoid tumors, where SWI/SNF factors critical for nucleosome movement needed to allow gene activation are defective (38). This has led to a clinical trial of the EZH2 Inhibitor Tazemetostat in Pediatric Subjects With Relapsed or Refractory INI1-Negative Tumors or Synovial Sarcoma (NCT02601937). EZH2 inhibition could also be useful in reactivating genes aberrantly suppressed in lymphoma due to the action of gain of function mutations of EZH2 (40–42) or due to redistribution of EZH2 due to overexpression of MMSET/NSD2 in t(4:14) myeloma (43). In our UTX wt/null (ARP-1/ARD) cell system, we found that UTX-null ARD cells were more sensitive than UTX-wt ARP-1 cells. Emerging data from our group using newly created UTX null cells by the Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) technique shows the same increased sensitivity of UTX null cells to EZH2 inhibitors. Furthermore, a recent publication showed that deletion of UTX in bladder cells using CRISPR increased the sensitivity of these cells to EZH2 inhibitor treatment (44). Collectively, these data show that UTX loss alters gene expression in MM cells potentially increasing their biological activity, contributing to tumor progression. However, UTX loss may represent an “Achilles Heel” of the tumor in that it appears to sensitize the cells to newly created EZH2 inhibitors. These data should stimulate the testing of EZH2 inhibitors in MM and other tumor displaying loss of UTX.

DISCUSSION

Hromas, Gainesville: Jon, have you ever looked at histone 3 lysine residue 36 trimethylation events in the cell lines? Because, as you know, dimethylation may simply be a stepping stone towards SETD2 mediated trimethylation.

Licht, Gainesville: We’ve measured histone methylation by mass spectroscopy, so we are not subject to the vagaries of antibody cross-reactivity, and there’s virtually no change in histone 3 lysine 36 trimethylation in the presence of overexpressed NSD2. Also, we found that UTX/KDM6A does not change methylation levels by the mass spectrometry assay.

The NSD2 enzyme system works at promoters and 5’ start sites of transcription of genes and seems to activate the promoter. SETD2-mediated histone 3 lysine 36 trimethylation is found in gene bodies, [it] prevents, helps attract DNA damage repair proteins, allows proper splicing to occur, and prevents aberrant re-recruitment of RNA polymerase into the body of genes. Recent studies show when you knockout SETD2 and you lose that K36 trimethylation, you get all kinds of abnormal transcripts starting in the middle of the gene and really is messing things up so it’s a very different system.

Hromas, Gainesville: SETD2 anomalies would also disrupt replication, for instance the trimethyl mark at replication forks. It seems to code for repair.