Advertisement

Chromatin modifications induced by PML-RARα repress critical targets in leukemogenesis as analyzed by ChIP-Chip

Claudia Hoemme, Abdul Peerzada, Gerhard Behre, Yipeng Wang, Michael McClelland, Kay Nieselt, Matthias Zschunke, Christine Disselhoff, Shuchi Agrawal, Fabienne Isken, Nicola Tidow, Wolfgang E. Berdel, Hubert Serve, Carsten Müller-Tidow

Abstract

The translocation t(15;17) generates the chimeric PML-RARα transcription factor that is the initiating event of acute promyelocytic leukemia. A global view of PML-RARα transcriptional functions was obtained by genome-wide binding and chromatin modification analyses combined with genome-wide expression data. Chromatin immunoprecipitation (ChIP)–chip experiments identified 372 direct genomic PML-RARα targets. A subset of these was confirmed in primary acute promyelocytic leukemia. Direct PML-RARα targets include regulators of global transcriptional programs as well as critical regulatory genes for basic cellular functions such as cell-cycle control and apoptosis. PML-RARα binding universally led to HDAC1 recruitment, loss of histone H3 acetylation, increased tri-methylation of histone H3 lysine 9, and unexpectedly increased trimethylation of histone H3 lysine 4. The binding of PML-RARα to target promoters and the resulting histone modifications resulted in mRNA repression of functionally relevant genes. Taken together, our results reveal that the transcription factor PML-RARα regulates key cancer-related genes and pathways by inducing a repressed chromatin formation on its direct genomic target genes.

Introduction

Balanced translocations that lead to expression of aberrant fusion proteins are a hallmark of acute leukemias.1 Many of these fusion proteins function as aberrant transcription factors that are an initiating event in leukemogenesis.2 Some translocations, eg those involving the MLL gene can lead to phenotypically diverse forms of leukemia, whereas the t(15;17) uniformly leads to acute promyelocytic leukemia. The reasons why some translocations are leukemia-type specific, whereas others are not, is unknown.

Histone modifications such as methylation of histone H3 at lysine 4 and 9 and acetylation of histone H3 are closely linked to the transcriptional activation status. In t(15;17), the chimeric PML-RARα fusion protein has been shown to recruit corepressors such as DAXX, histone deacetylase (HDAC) activity, DNA methyltransferase activity, and the SUV39H1 histone methyltransferase to RARβ2, the most extensively studied target gene.37 PML-RARα also interacts with histone deacetylase 1.8 PML-RARα homodimerization has been shown to relax the relatively stringent RARα DNA binding specificity.9,10 This gain of function is supposed to lead to many additional genomic binding sites that are not well defined. As a consequence, virtually all direct genomic targets of PML-RARα are currently unknown. In addition, on a global level, the mechanistic alterations occurring at PML-RARα target genes remain to be clarified.

On a phenotypic level, the PML-RARα fusion protein blocks differentiation and apoptosis and enhances self-renewal.11,12 In mouse models, PML-RARα induces a disease similar to acute promyelocytic leukemia (APL).13,14 Microarray analyses elucidated several leukemogenic mechanisms and pathways.1519 For example, PML-RARα induces activation of the Wnt signaling pathway.17 Also, PML-RARα alters the apoptotic response and expression of differentiation genes.15 However, these studies do not distinguish between direct and indirect effects on gene expression. This knowledge is crucial to understand the mechanistic implications of PML-RARα and to elucidate the reasons for the unique phenotype that is associated with its activities.

The possibilities to understand transcription factor functions has recently been significantly improved by genome-wide approaches that identify target genes in vivo using Chromatin immunoprecipitation (ChIP)–Chip approaches.20,21 In addition, the ability to use this method to map epigenetic modifications such as histone acetylation at promoters known to bind the transcription factor allows for the identification of the functional consequences of transcription factor binding to its genomic targets.22,23

Using ChIP-chip analyses, we identified 372 direct PML-RARα genomic targets and show that PML-RARα induces heterochromatin formation on virtually all of its identified target genes. Several of the identified genes are known tumor suppressors and for one of the novel genes (S100P), we show a potential role in the PML-RARα–associated block in differentiation. A subset of target genes was confirmed in primary leukemic blasts from a patient with APL, indicating that these genes may play a role in acute promyelocytic leukemia.

Methods

The use of human material for scientific purpose was approved by the ethics committee of the University of Münster. Informed consent was obtained in accordance with the Declaration of Helsinki.

Cell culture and patient sample

U937 cells were cultured in RPMI with l-glutamine and 10% fetal calf serum. The stably transfected U937 cell lines that express PML-RARα or empty vector control in a Zn2+-inducible fashion have been described previously.11

Prior to ChIP or RNA isolation, U937 cells were exposed to 0.1 μM 5-aza-2′-deoxycytidine (5-aza-dC) for 6 days to allow the analysis of histone modifications with a reduction of the density of a priori DNA-methylated genes. Expression of the transgene was subsequently induced by addition of 0.1 mM ZnSO4 for the indicated times. Density centrifugation–enriched primary blasts were obtained from a patient with newly diagnosed APL and t(15;17) after informed consent. The primary APL cells were formaldehyde fixed and stored at −80°C until used. The patient's bone marrow showed a classical FAB-M3 phenotype, and the patient achieved complete remission after treatment with all-trans retinoic acid (ATRA) and chemotherapy (daunorubicin and cytarabine).

ChIP

ChIPs were performed essentially as described.17 Briefly, 2 × 107 cells were fixed with formaldehyde, neutralized with glycine, and rinsed with cold phosphatate-buffered saline. After lysis of the cells, samples were sonicated to an average DNA length of 500 bp. Immunoprecipitation of 0.5 mg precleared chromation was carried out by addition of 3 μg of the following antibodies: α-PML (Santa Cruz Biotechnology, Santa Cruz, CA), α-acetyl H3, α-dimethyl H3 (Lys4), α-trimethyl H3 (Lys4), α-dimethyl H3 (Lys9), α-trimethyl H3 (Lys9), α-HDAC, and α-histone H3 (Upstate Biotechnology, Charlottesville, VA). Another α-trimethyl H3 (Lys4) antibody (Abcam, Cambridge, MA) was used to confirm the findings. At least 2 independent ChIPs were performed for each cell line and antibody.

DNA amplification and labeling

Chromatin-immunoprecipitated DNA was amplified in 2 steps, including a T7 sequenase extension using random primer with a fixed sequence linker and a second step of amplification using the fixed sequence primer and Taq polymerase. The products were purified and labeled with amino-allyl–conjugated dUTP using the BioPrime labeling kit (Invitrogen, Carlsbad, CA) and random primer. Products were purified and crosslinked with monofunctional NHS-ester Cy3 or Cy5 dye.

Microarray hybridization

The promoter arrays with about 12 000 human promoters spotted in triplicates have been described24 (http://bioinformatics.skcc.org/array). CpG Island arrays containing about 12 000 CpG island clones derived from the Sanger Center (Cambridge, United Kingdom) were purchased from the University Health Network (UHN) Microarray Centre (Toronto, ON).25 U937 genomic DNA (input control) and amplified ChIPs were labeled with Cy5 and Cy3, respectively. Hybridization and washing of the arrays was performed as described.24 In total, 26 promoter arrays and 46 CpG island arrays were included in the final analysis. Raw data and Excel sheets of all arrays will be available on our website and submitted to the GEO database (National Center for Biotechnology Information).26

Data analysis

Microarrays were scanned and images were analyzed with Spotreader software (Niles Scientific, Portola Valley, CA). Spots with aberrant morphology or those with intensities below the threshold of detection were flagged. Cy3/Cy5 ratios were print-tip Lowess normalized using the Diagnosis and Normalization of Microarray Data (DNMAD) tool (http://dnmad.bioinfo.cnio.es/). Subsequently, the arrays were median centered. Class comparisons at a P value less than .01 were performed with the BRB Array Tools software (National Cancer Institute, http://linus.nci.nih.gov/BRB-ArrayTools.html). To account for differences in nucleosome density, the ratios between U937-PML-RARα and empty vector expressing cell lines were standardized using the ratios of all analyzed histone modifications in this study (AcH3, di-mH3K4, tri-mH3K4, di-mH3K9, and tri-mH3K9). Visualization of enriched PML-RARα–bound chromosomal loci was performed with Mayday (Center for Bioinformatics, Tübingen, Germany).27 The Clover software was used to identify the statistically overrepresented retinoic acid response elements (RAREs) in the PML-RARα target promoters.28,29 The Clover program (Bioinformatics Program, Boston University, Boston, MA) calculates a similarity score between a weight matrix and a position in a DNA sequence, taking into account a set of background sequences. For the promoter sequences we used a score threshold of 6.0 and a P value less than .05. For the CpG island sequences, no hits were found with these values; therefore, we had to enlarge the P value to .3.

Quantitative real-time PCR

Quantitative real-time polymerase chain reaction (PCR) using Sybr Green (Applied Biosystems, Foster City, CA) was performed as described to verify ChIP-Chip microarray analysis.17 In addition to cell-line chromatin, we used chromatin of a patient with newly diagnosed APL and t(15;17). PCR amplicons of the genomic regions were designed using the Primer Express software (Applied Biosystems), and primer sequences are provided in Table S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article). To account for different amounts of immunoprecipitated chromatin, the Ct values were standardized to Ct values of the 10 000 bp upstream primer of p21CIP1 (ΔΔCt method), which we found to be unaltered in transcription factor binding or histone modifications in our experiments.

Gene-expression analysis

Gene-expression analysis was performed with Affymetrix high-density oligonucleotide microarrays (GeneChip HG-U133A, Affymetrix, Santa Clara, CA) as described.17

A class comparison of the data were computed using the BRB Array Tools software. Differentially regulated genes were hierarchically clustered and visualized using Mayday.27

Quantitative real-time PCR was performed using Sybr Green technology (Applied Biosystems) to verify mRNA expression of PML-RARα target genes.17

Electroporation and differentiation of U937 cells

A total of 107 zinc-inducible U937-PML-RARα or U937–empty vector cells were electroporated with pCDNA3-GFP or pCDNA3-GFP-S100P with an EPI 2500 gene pulser (Fischer, Heidelberg, Germany) and exposed to 0.1 mM ZnSO4. Cells were treated with 10−7 M vitamin D3 24 hours after electroporation and stained with α-CD11b-PE or α-CD14-PE antibodies. Expression of CD11b and CD14 was analyzed by flow cytometry in green fluorescent protein–positive (GFP+) cells after 48 hours of vitamin D3 exposure.

Western blotting and kinase assays

Western blot analyses were performed as described.17 Antibodies against CDK2 and CDK4 were obtained from Santa Cruz Biotechnology, the anti-RGS2 antibody was from Abnova (Heidelberg, Germany), and the anti-ANKRD2 antibody was a kind gift from Dr G. Faulkner (International Centre of Genetic Engineering and Biotechnology, Trieste, Italy). Kinase assays were performed as described.30

Results

Identification of PML-RARα genomic targets by ChIP-chip

ChIP was performed with α-PML antibodies to enrich for PML-RARα–bound genomic sequences in zinc-induced U937-PML-RARα cells and as a negative control in zinc exposed U937–empty vector cells.11 All comparisons in the cell lines were made between PML-RARα–expressing and –nonexpressing cells. DNA sequences specifically precipitated by α-PML antibody in PML-RARα–expressing cells (but not in PML-RARα cells) most likely present PML-RARα–specific targets. ChIPs were linearly amplified and subsequently hybridized to promoter and CpG island arrays. At least 2 independent ChIPs were hybridized to the genomic arrays. Overall, ChIP hybridizations (PML-RARα binding, HDAC1 binding, and methylation of histone H3) were analyzed from 26 promoter array sets and 46 CpG island arrays. Following intraarray and interarray normalization, PML-RARα–bound sequences were identified by applying a class comparison (U937-PML-RARα versus U937–empty vector cells) for each array set using BRB Array tools. Overall, 372 genomic locations with significant PML-RARα binding were identified (Table S2). In the promoter arrays, about 50% of the enriched sequences contained bona fide RAREs (each P < .05). In the CpG islands only about 30% of the targets contained RAREs (P < .3). Overall, only about 40% of the presumed genomic targets do contain classical RAREs. These findings are consistent with the relaxed and altered DNA-binding properties of PML-RARα compared with RARα and confirms the necessity to specifically identify PML-RARα targets. The presence and locations of the predicted RAREs are indicated in Tables 1 and S2.

View this table:
Table 1

PML-RARα promoter occupancy

PML-RARα binding is associated with chromatin modifications

Chromatin modifications of PML-RARα–bound promoters can indicate the functional consequences such as transcriptional repression. PML-RARα binding has previously been shown to lead to recruitment of HDACs and the SUV39H1 methyltransferase at the RARβ2 gene.3,4 Using ChIP-Chip, we focused on several modifications of histone H3 (AcH3, di-mK4, tri-mK4, di-mK9, and tri-mK9) in the same cell lines that were used for PML-RARα target identification. These analyses led to the identification of 1800 promoters and CpG islands that differed between PML-RARα–expressing and –nonexpressing cells in histone modification and/or PML-RARα binding. These data contain direct PML-RARα effects as well as indirect and secondary changes.

Heatmaps of PML-RARα–bound promoters and CpG islands as well as histone modifications are shown in Figure 1A and B, respectively.

Figure 1

PML-RARα binding and histone modification changes. ChIPs were performed with zinc-induced U937-PML-RARα cells (Pr9) and U937-empty vecor cells (PMT) with the indicated antibodies. Amplified ChIPs were hybridized on promoter and CpG island microarrays. Bioinformatics analyses identified PML-RARα–bound sequences and the histone modifications in the 2 cell lines. (A,B) The heatmaps of the 372 target genes indicate the average ratio between Pr9 versus PMT of the different arrays performed for each antibody. All values are visualized using the log ratio of signals in U937-PML-RARα versus control cell line (averages of 2-4 independent experiments). Red indicates an enrichment and green a decrease for PML-RARα or HDAC binding or the respective histone modification. The genes were clustered using a self-organizing map algorithm (with a 2 × 2 grid). The heatmaps were computed with Mayday. (A) Heatmap of promoter loci (271 genomic PML-RARα targets). (B) Heatmap of CpG island loci (101 genomic PML-RARα targets). (C-E) Enrichment of PML-RARα binding was categorized as absent (no), weak (> 1.25), intermediate (> 1.75), and strong (> 2.25) enrichment. The box plots indicate the range of changes for the different histone modifications between the promoters in U937-PML-RARα expressing versus U937–empty vector cells. The boxes indicate the range of 50% of the data points, and the whiskers contain 75% of the observed data points. The medians are indicated as horizontal lines in the boxes. The changes for acetylation of histone H3 and trimethylation of lysine 9 are statistically significant (P < .01), whereas dimethylation of lysine 4 is not.

PML-RARα–occupied promoters showed a significant decrease in histone H3 acetylation and increase in lysine 9 trimethylation (each P < .001; Figure 1C,D). In contrast, dimethylation of histone H3 at lysine 4 was not affected (Figure 1E). These findings indicate that most of the genes with PML-RARα enrichment in our analysis are likely to be true targets in vivo and second, that PML-RARα exerts epigenetic alterations on most or all of its target genes.

The changes in HDAC1 binding and histone modifications were analyzed separately for PML-RARα–bound and –nonbound genomic locations. This way, we were able to further distinguish between the direct and indirect effects. Figure 2 shows the changes between PML-RARα–expressing and –nonexpressing cells on the right-hand side for direct PML-RARα targets and on the left for other genomic sequences altered in histone modification patterns upon PML-RARα expression. PML-RARα binding increased trimethylation of lysine 4 for most of its target genes, which inversely correlated with dimethylation of lysine 4 (Figure 2A). Trimethylation of lysine 4 usually indicates chromatin primed to undergo transcription,31,32 but might also play a role in active gene suppression.33 Interestingly, increases in H3-K9 trimethylation and histone acetylation were mutually exclusive in PML-RARα–bound and –nonbound sequences (Figure 2B). Virtually all PML-RARα target genes exhibited increased H3-K9 trimethylation and a decrease in histone H3 acetylation. The loss in histone H3 acetylation was closely associated with increased HDAC1 binding (Figure 2C). These changes are consistent with induction of a repressive chromatin state by PML-RARα in almost all of its direct target genes.

Figure 2

Identification of PML-RARα–associated chromatin modifications. The dot plots on the right indicate the changes in histone modifications and HDAC1 binding in promoters bound by PML-RARα. To allow comparisons, the panels on the left show promoters with altered histone modifications that are not bound by PML-RARα. Each dot identifies one promoter or CpG island. Data are based on the promoter arrays except the plots involving HDAC1 binding, which are based on the CpG island arrays. All axes indicate the difference between zinc-induced PML-RARα–expressing cells (Pr9) versus control cells (PMT). The indicated correlation coefficients are all significant at the P value less than .001 level, with the exception of r = −.18 (P = .03) for the association between trimethylation of lysine 9 and histone H3 acetylation for PML-RARα–bound genes. (A) Association between trimethylation of H3 lysine 4 and dimethylation of H3 lysine 4. Quadrant i shows an increase in trimethylation and a decrease in dimethylation of histone H3 lysine 4 for PML-RARα bound promoters. (B) Association between H3 lysine 9 trimethylation and H3 acetylation. Quadrant iv indicates a decrease in H3 acetylation and an increase in H3 lysine 9 methylation in most direct PML-RARα target genes (right panel), but not in promoters that are not targeted by PML-RARα (left panel). (C) Association between H3 acetylation and HDAC1 binding. Quadrant iv shows an decrease in H3 acetylation and an increase in HDAC1 binding.

Verification of PML-RARα targets and chromatin modifications by real-time PCR

ChIP coupled with real-time PCR was used to verify the microarray results. PML-RARα binding was verified for 20 of 23 tested genomic localizations in vivo (Table 1). Most of these also showed increases in HDAC1 binding and H3K4 and H3K9 trimethylation as well as loss in H3 acetylation levels (Figure 3A,C). Since the increase in trimethylation of histone H3 lysine 4 (in ChIP-Chip and ChIP-PCR assays) was unexpected for PML-RARα targets, we repeated the experiments with a second antibody against H3K4me3 for selected target genes (Figure 3B). Again, these analyses verified a significant increase in H3K4 trimethylation at PML-RARα genomic targets.

Figure 3

Verification of PML-RARα targets in U937 cells and in primary patient blasts. (A) ChIPs of zinc-induced U937-PML-RARα cells and U937–empty vector cells were performed with the indicated antibodies. Changes in PML-RARα binding and histone modifications for selected targets previously identified by microarrays were verified by real-time PCR. The fold change of PML-RARα or HDAC1 binding and the histone modifications in U937-PML-RARα cells versus U937-control cells are indicated. (B) A total of 2 different antibodies were used to verify enrichment of H3K4 trimethylation at PML-RARα–bound promoters by real-time PCR. The increase in H3K4 trimethylation in PML-RARα–expressing U937 cells versus U937 control cells is indicated for selected target genes. (C) PML-RARα binding was verified for selected targets by real-time PCR in U937-PML-RARα cells. The ratio between U937-PML-RARα cells versus U937-control cells is indicated for the ChIP-Chip assays and the PCR confirmation. (D) Verification of PML-RARα binding in primary patient APL. The fold enrichment in PML-RARα binding was analyzed by ChIP with an α-PML antibody versus IgG isotype control antibody.

We also used ChIP–real-time PCR to verify PML-RARα targets in primary patient cells. Anti-PML ChIPs were carried out, and results were compared with ChIPs with control IgG. A total of 13 of the 20 verified PML-RARα targets in U937 cells were confirmed in primary blasts from a patient with APL (Table 1; Figure 3D). Although this approach does not readily exclude that normal PML (but not PML-RARα) may be localized at the respective promoters, this scenario is unlikely given that PML-RARα specifically binds to these sequences in U937-PML-RARα cells. These confirmatory analyses suggested that the microarray analyses revealed true targets that are occupied by PML-RARα in primary APL.

PML-RARα occupancy was analyzed by real-time PCR in more detail at 2 target gene promoters, DEF6 and p21CIP1. In U937 cells as well as in the primary patient sample, discrete regions for PML-RARα binding were identified that matched between the cell lines and the patient leukemia cells (Figure 4A,B). For the p21CIP1 promoter, a RARE has been described about 1200 bp upstream of the transcriptional start. This site matched the strong enrichment of PML-RARα observed in our analyses.34 The Clover software predicted a second RARE at −270 bp upstream of the transcriptional start site, which was associated with a second peak for PML-RARα enrichment. The p21CIP1 mRNA was 5-fold repressed on the mRNA level (Figure 5) in U937-PML-RARα cells and significantly repressed upon PML-RARα expression on the protein level (Figure 4C). Functionally, CDK2 and CDK4 kinase activity was increased upon PML-RARα induction consistent with a repression of p21CIP1.

Figure 4

DEF6 and p21 are targets of PML-RARα. (A) ChIPs of zinc-induced U937-PML-RARα and U937–empty vector cells were performed with an α-PML antibody. The localization of PML-RARα binding at the DEF6 and p21 promoters was analyzed in U937 cells using a series of real-time PCR amplicons that tiled a 5-kb region surrounding the transcriptional start site. The fold enrichment of signal in PML-RARα–expressing U937 cells is indicated compared with empty vector–expressing U937 cells. (B) ChIPs with an α-PML antibody and an IgG control antibody were performed with blasts from a patient with APL with the t(15;17) translocation. The fold enrichment of signal in the ChIPs with α-PML antibody is shown compared with ChIPs with an IgG control antibody. (C) Top panel: Immunoblot analysis was performed for p21 from whole-cell lysates of U937-PML-RARα cells before and after PML-RARα induction. Bottom panel: CDK2 and CDK4 were immunoprecipitated before and after induction of PML-RARα. Histone H1 and GST-Rb were used as substrates for in vitro kinase reaction to analyze the kinase activity of CDK2 and CDK4 in PML-RARα–expressing cells. PML-RARα induction was associated with an increase in CDK2 and CDK4 kinase activity.

Figure 5

mRNA and functional analyses of PML-RARα targets. (A) Changes in mRNA expression of selected genes between zinc-induced U937-PML-RARα and U937–empty vector cells were confirmed by real-time RT-PCR. The bars indicate the fold mRNA repression upon PML-RARα expression for the microarray data and the real-time RT-PCR data (mean plus standard deviation). (B) Protein expression of ANKRD2 and RGS2 was decreased in U937-PML-RARα cells compared with U937–empty vector cells 12 hours after zinc induction as analyzed by Western blotting. (C) S100P expression can overcome the PML-RARα–induced differentiation block. U937-PML-RARα and U937–empty vector cells were transfected with GFP or GFP-S100P, subsequently zinc induced, and exposed to vitamin D3 24 hours after transfection. After 48 hours, cells were analyzed for CD11b and CD14 expression in GFP+ cells (mean plus standard deviation of 2 independent experiments).

PML-RARα promoter binding is associated with a decrease in gene expression

PML-RARα binding leads to chromatin changes, which are presumed to be associated with transcriptional repression. We therefore analyzed gene expression by high-density Affymetrix (HG-U133A) arrays after induction of PML-RARα in U937 cells compared with empty vector (PMT vector) U937 cells. Affymetrix IDs were matched to the promoter arrays based on genomic localization. All the promoter array clones are located directly upstream of the putative transcripts. On the CpG island array, a CpG island was regarded to be likely associated with a gene if the CpG island was located up to 20 000 bp upstream of the transcriptional start site. Almost all of the matched pairs of CpG islands and genes were located in proximity of about 1000 bp to each other. We compared gene expression differences between U937-PML-RARα and control cells for the identified PML-RARα target genes. U937 cells expressing PML-RARα showed significantly reduced mRNA expression levels of genes occupied by PML-RARα (Figure 6A). Besides PML-RARα binding, a significant association with mRNA expression changes was also found for histone H3 acetylation changes and for changes in trimethylation of H3 lysine 9 (Figure 6B,C). No significant alterations in gene expression levels were observed upon changes in dimethylation levels of histone H3 lysines 4 and 9 (data not shown).

Figure 6

PML-RARα–associated gene expression changes and regulated molecular functions. (A-C) Depiction of global alterations in mRNA expression with regard to PML-RARα binding or alterations in histone H3 acetylation or H3K9 trimethylation. The mRNA expression changes were analyzed by Affymetrix mRNA expression arrays (GeneChip HG-U133A) in 12-hour zinc-induced U937-PML-RARα cells versus U937–empty vector cells. (A) PML-RARα binding is associated with a significant decrease in mRNA expression levels. (B) Genes with increased histone H3 acetylation show increased gene expression and vice versa. (C) Increases in histone H3 lysine 9 trimethylation are associated with transcriptional repression. The boxes indicate the range of 50% of the data points and the whiskers contain 75% of the of the observed data points. (D) Analysis of the molecular functions of PML-RARα target genes. Significantly regulated genes for genomic PML-RARα binding, mRNA expression changes in PML-RARα–expressing U937 cells, and mRNA expression changes in PML-RARα–expressing murine promyelocytes (dataset kindly provided by Dr T. J. Ley) were analyzed for their association with molecular functions using Ingenuity IPA software 3.0. Molecular functions with significant (P < .05) enrichment in each of the 3 datasets are indicated.

Identification of molecular functions regulated by PML-RARα on genome and transcriptome levels

Additional experiments were performed to verify whether PML-RARα promoter occupancy is associated with mRNA expression changes. To analyze alterations induced by PML-RARα at the transcriptome level, we performed class comparison analysis of the Affymetrix microarray data. Overall, the mRNA expression of 715 genes differed significantly between U937-PMT and U937-PML-RARα at the P value less than .01 level (Table S3; Figure S1).

To extend the PML-RARα–associated mRNA expression changes beyond the human cell line model, we used mRNA expression changes from a transgenic mouse model as described by Walter et al.19 The original raw data were kindly provided by Dr T. J. Ley (Washington University, St Louis, MO). In this approach, promyelocytes were isolated from PML-RARα transgenic mice with no leukemic phenotype as well as from control littermates. Class comparison analyses indicated 170 significantly different genes at the P value less than .01 level (Table S4).

The lists of significant changes in U937 cells (genomic and transcriptomic changes) and the murine promyelocyte data were loaded into the Ingenuity IPA 3.0 software (Ingenuity, Redwood City, CA). The lists of significant changes in U937 cells and the murine promyelocyte data were analyzed by a pathway analysis software (Ingenuity IPA 3.0; Ingenuity). Among the molecular functions examined, PML-RARα–regulated genes contained a significant enrichment for genes involved in cell cycle, development, cellular growth and proliferation, DNA repair, and cell death. These molecular functions were independently and significantly enriched at the P value less than .05 level in each of the 3 datasets (Figure 6D).

Funtional relevance of the identified PML-RARα targets

The regulation of several identified PML-RARα targets was confirmed at the mRNA expression level by real-time PCR. A total of 8 of 10 selected genes showed a transcriptional repression in U937-PML-RARα cells compared with U937-control cells (Figure 5A). The most strongly repressed genes were S100P, ANKRD2, and RGS2. The suppression of ANKRD2 and RGS2 on the protein level is shown in Figure 5B (endogenous S100P was not detectable with the available antibodies). RGS2 is a very interesting novel PML-RARα target, since we have recently shown that RGS2 is a potential tumor suppressor in acute myeloid leukemia (AML).35 RGS2 is repressed by Flt3 mutations and regulates C/EBPα expression. Importantly, it promotes granulocytic differentiation. Another gene that is involved in myeloid differentiation is S100P.36 Therefore, we analyzed the potential of S100P to overcome the PML-RARα–induced differentiation block. U937–empty vector cells (PMT) or U937-PML-RARα cells (Pr9) were transfected with GFP- or GFP-S100P–expressing plasmids and exposed to zinc. Vitamin D3 was added 24 hours after transfection, and cells were analyzed for differentiation 48 hours later. PML-RARα inhibited differentiation of vitamin D3–exposed U937 cells as described.11 However, expression of S100P but not GFP at least partially overcame the differentiation block as determined by the increase in CD11b and CD14 expression (Figure 5C).

Discussion

PML-RARα is a potent oncogene that induces APL in humans and in various mouse models. This is the first study to identify its genomic targets and the resulting chromatin modifications on a global level. Several important findings were obtained: PML-RARα acts universally as a suppressor at its occupied promoters. PML-RARα binding to its target genes was accompanied by HDAC1 recruitment, loss of histone H3 acetylation, and increased trimethylation of histone H3 lysine 9. These findings are in line with reports of a repressive chromatin structure found at the RARβ2 locus upon PML-RARα expression. PML-RARα recruits corepressors and HDAC activity to its targets.3 Methylation of H3 lysine 9 is likely due to recruitment of the SUV39H1 methyltransferase that catalyzes this reaction.4 Unexpectedly, our analyses also indicated an increase in trimethylation of histone H3 lysine 4 concomitant with a decrease in dimethylation of the same lysine. This finding suggests that PML-RARα induced a switch from the dimethylated toward the trimethylated form of lysine 4. Nuclear receptors such as RARα can recruit H3K4 methyltransferases.37 Our findings at a large number of direct targets suggest that PML-RARα retains the ability to recruit H3K4 methyltransferase activity in the absence of retinoic acid. H3K4 methylation often is associated with promoters and 5′ untranslated regions (UTRs) of actively transcribed genes.32,38 However, recent results suggest that the functional relevance of H3K4 trimethylation depends on the recruited effector molecules. For example, H3 lysine 4 methylation can function in active gene repression by recruiting PHD-containing transcriptional repressors.33 Another nuclear receptor, the androgen receptor, has recently also been shown to induce lysine 4 methylation concomitant with transcriptional repression.39

The direct PML-RARα genomic target genes belong to diverse groups of genes with critical functions in cellular processes. The wide range of target promoters might explain the pleiotropic effects of PML-RARα. These include several master regulators for transcriptional programs such as HOX genes (HOXD13), histone acetyltransferase HBO-1 (MYST2), and other transcriptional regulators (SREBF-1 and PPARD). For example, HOXD13 is a target for balanced translocations in AML itself,40 and the Nup98-HOXD13 fusion protein induces an AML-like disease alone and in combination with known leukemia-inducing genes.41,42 In addition, PML-RARα also directly binds to the promoter of the CDKN1A/p21CIP1 gene that regulates cell-cycle progression and stem cell functions.43,44 Previously, CDKN1A/p21 has been described to be induced by PML-RARα upon exposure to ATRA.45 Nonetheless, our data clearly show that PML-RARα binds to the p21 promoter. HDAC1 is recruited and histone H3 acetylation is decreased at the p21 promoter. In line with these data, we find reduced p21 mRNA expression levels in our expression microarrays and by real-time reverse transcription (RT)–PCR. Our analyses suggest that p21, like most other direct targets, is repressed by PML-RARα on the transcriptional level. The discrepancy to the published data highlights the fact that genes are regulated at multiple levels and that indirect effects occur that might in specific situations (eg, by addition of ATRA) overshadow the direct genomic effects.

Several other direct PML-RARα targets are known to be closely involved in cell-cycle regulation, cellular differentiation, and tumor suppression. The ANKRD2 promoter is bound by PML-RARα and ANKRD2 mRNA, and protein is repressed by PML-RARα. ANKRD2 interacts in vitro and in vivo with p53 and enhances the induction of the p21 promoter by p53.46 Loss of ANKRD2 could further reduce p21 promoter activity. Several of the identified target genes might play a role in leukemogenesis. For example, RGS2 is a regulator of G-protein signaling that we recently identified as a potential tumor suppressor in AML.35 Its expression is strongly inhibited by Flt3 mutations that often cooperate with PML-RARα. RGS2 also regulates C/EBPα expression, whose function is known to be altered in APL. Our finding that RGS2 is a direct genomic target of PML-RARα and that RGS2 mRNA and protein are subsequently repressed highlights the potential importance of this gene in AML and especially in APL pathogenesis. As a proof of principle for the potential significance of other genomic targets, we used S100P, a small Ca2+-binding protein involved in migration and potentially metastasis.47 The S100P promoter was bound by PML-RARα, and mRNA expression was significantly reduced. Native protein expression was not detectable with available antibodies (data not shown), which is similar to the situation in lung cancer where changes in mRNA expression are still associated with altered metastatic activity of lung cancer cells.47 S100P has recently been shown to be involved in differentiation of human myeloid leukemia cells.36 In our experiments, expression of S100P in U937-PML-RARα cells overcame the differentiation block. These findings suggest that S100P may play a role in the PML-RARα–induced differentiation block.

Taken together, the diverse range of direct targets might explain the unique features of APL and the specific type of leukemia induced by PML-RARα. The unparalleled effectiveness of retinoic acid–based differentiation therapy in PML-RARα–associated APL48 is likely to depend on the leukemogenic relevance of its direct genomic targets.

The identification of the direct PML-RARα targets should help to better understand the pathogenesis of APL and could aid in devising improved differentiation therapies in other leukemias.

Figure S1

Supplementary PDF file available online.

Table S1

Supplementary PDF file available online.

Table S2

Supplementary PDF file available online.

Table S3

Supplementary PDF file available online.

Table S4

Supplementary PDF file available online.

Authorship

Contribution: C.H. and C.M.-T. designed research, performed research, analyzed data, and wrote the paper; A.P., C.D., S.A., F.I., and N.T. performed research; G.B., M.McC., and W.E.B. wrote the paper; Y.W. contributed new reagents; K.N. and M.Z. analyzed data; and H.S. designed research and wrote the paper.

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

Correspondence: Carsten Müller-Tidow, Department of Medicine A, Hematology and Oncology, University of Münster, Domagkstr. 3, 48149 Münster, Germany; e-mail: muellerc{at}uni-muenster.de.

Acknowledgments

We are grateful for excellent technical assistance to Sandra Dohts and Barbara Mlody. We thank the Biometric Research Branch (BRB) Array tools team and especially Supryia Menendez for help with the software. We thank Dr T. J. Ley for providing the transgenic murine promyelocyte dataset. We thank Xiao-Qin Xia for the promoter array manufacture and Fred Long for bioinformatics support. We are grateful to Dr Faulkner for providing anti-ANKRD2 antibody. We are very grateful to Dan Mercola and Shilpi Arora for all their work and efforts with the microarray production.

This work was supported by the NGFN-2 LeukemiaNet, the José-Carreras Leukämiestiftung, the Deutsche Forschungsgemeinschaft, and the Deutsche Krebshilfe. C.M.-T. was supported by a Heisenberg grant from the Deutsche Forschungsgemeinschaft. M.M. was supported by National Institutes of Health grants R01CA68822 and U01CA0114810 and Department of Defense grants DAMD81-06-1-0253 and DAMD17-03-1-0022. K.N. and M.Z. were supported by the Deutsche Forschungs-gemeinschaft, AZ BIZ 1/1-3. Their work was supported by National Institutes of Health grant CA 114810.

R01CA68822U01CA0114810CA 114810National Institutes of Health

Footnotes

  • An Inside Blood analysis of this article appears at the front of this issue.

  • The online version of this article contains a data supplement.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

  • Submitted March 14, 2007.
  • Accepted November 5, 2007.

References

View Abstract