RARα-PLZF overcomes PLZF-mediated repression of CRABPI, contributing to retinoid resistance in t(11;17) acute promyelocytic leukemia

Abstract

Leukemia-associated chimeric oncoproteins often act as transcriptional repressors, targeting promoters of master genes involved in hematopoiesis. We show that CRABPI (encoding cellular retinoic acid binding protein I) is a target of PLZF, which is fused to RARα by the t(11;17)(q23;q21) translocation associated with retinoic acid (RA)-resistant acute promyelocytic leukemia (APL). PLZF represses the CRABPI locus through propagation of chromatin condensation from a remote intronic binding element culminating in silencing of the promoter. Although the canonical, PLZF-RARα oncoprotein has no impact on PLZF-mediated repression, the reciprocal translocation product RARα-PLZF binds to this remote binding site, recruiting p300, inducing promoter hypomethylation and CRABPI gene up-regulation. In line with these observations, RA-resistant murine PLZF/RARα+RARα/PLZF APL blasts express much higher levels of CRABPI than standard RA-sensitive PML/RARα APL. RARα-PLZF confers RA resistance to a retinoid-sensitive acute myeloid leukemia (AML) cell line in a CRABPI-dependent fashion. This study supports an active role for PLZF and RARα-PLZF in leukemogenesis, identifies up-regulation of CRABPI as a mechanism contributing to retinoid resistance, and reveals the ability of the reciprocal fusion gene products to mediate distinct epigenetic effects contributing to the leukemic phenotype.

Acute myeloid leukemia (AML) is highly heterogeneous at the molecular level, and to date chromosomal breakpoint junctions have been cloned from >100 reciprocal balanced translocations associated with the disease (http://cgap.nci.nih.gov/Chromosomes/Mitelman). These lead to the formation of chimeric oncoproteins, which play a critical role in leukemogenesis. However, despite the apparent diversity of genes implicated in AML, some recurring mechanistic themes have emerged. In particular, chromosomal translocations commonly involve genes encoding hematopoietic transcription factors such as RUNX1 (AML1) and retinoic acid receptor α (RARα), generating fusion proteins that recruit corepressor complexes including histone deacetylases (HDACs) and DNA methyltransferases (DNMTs), leading to targeted repression of genes implicated in myeloid differentiation (reviewed in refs. 1–3). Although such mechanisms could feasibly induce the differentiation block that characterizes AML, it is clear that this is insufficient for the full leukemic phenotype (3). Therefore, a key question is the relative contribution of the “fusion partner” moiety of chimeric oncoproteins that involve hematopoietic transcription factors. In some instances, the translocation partner serves to recruit corepressor complexes, as exemplified by ETO and TEL (ETV6) that are fused to AML1 in AML and acute lymphoblastic leukemia, respectively (1, 4). A further common feature is the provision of a dimerization interface, as is the case for each of the six potential fusion partners [PML, PLZF, NPM1, NuMA, PRKAR1A (63), and STAT5b] of RARα in acute promyelocytic leukemia (APL) (reviewed in refs. 5 and 6). Although the resultant forced dimerization of RARα has been shown to contribute to the transforming function of APL fusion proteins (7–9), studies undertaken using transgenic mouse models have highlighted the importance of the fusion partner to the pathogenesis of the disease (10–12). Indeed, these have a critical bearing on the disease phenotype, particularly with respect to sensitivity to all-trans-retinoic acid (ATRA) (reviewed in ref. 13). Cases involving PML, NPM1, and NuMA as the fusion partner respond clinically to this agent (14–16), whereas APL involving PLZF (promyelocytic leukemia zinc finger), due to the t(11;17)(q23;q21) chromosomal translocation, is typified by retinoid resistance and a poorer prognosis (17). Previous studies have linked this resistance to retinoid-insensitive binding of corepressor complexes to the N-terminal POZ repressor domain of PLZF, which is retained in the PLZF-RARα fusion protein (18–21). However, there is evidence to suggest that the reciprocal RARα-PLZF chimeric protein plays a significant role in mediating the retinoid resistance that characterizes this subset of APL, although the mechanism had hitherto been uncharacterized. Blasts from transgenic mice coexpressing PLZF-RARA and RARA-PLZF fusion transcripts have been shown to be substantially more resistant to differentiation with retinoic acid (RA) than cells expressing PLZF-RARA alone (22). Similarly, APL cases in which both fusion gene products were expressed exhibited primary resistance to RA (17, 23–25), whereas blasts from a case in which PLZF-RARA was the sole fusion transcript formed as a result of a cryptic insertion event [rather than the t(11;17) balanced translocation] were sensitive to ATRA in vitro, correlating with achievement of clinical remission with this agent (26).

Given that there is evidence suggesting that the reciprocal RARα-PLZF fusion protein is not only implicated in mediating retinoid resistance but also cooperates with PLZF-RARα in APL pathogenesis (22), the identification of cases in which the reciprocal RARA-PLZF fusion gene is not formed because of nonreciprocal chromosomal rearrangements (25), provides an opportunity to identify downstream target genes, and further dissect processes underlying the leukemic phenotype. Previous studies have identified cyclin A2 (27), c-Myc (28), and HoxB2 (29) as RARα-PLZF targets; here we identified the gene encoding cellular retinoic acid-binding protein I (CRABPI) (30), which is involved in cellular catabolism of retinoids (31, 32) and is a well established mediator of retinoid resistance in a variety of cellular systems (33–37), as a target of RARα-PLZF. We have characterized the mechanisms by which PLZF functions as a long-range repressor of the CRABPI locus and demonstrated that the RARα-PLZF protein acts as a dominant positive regulator, binding and altering transcription of PLZF target genes.

Results

Identification of CRABPI as a Potential Downstream Target of RARα-PLZF in t(11;17)-Associated APL.

To examine the role of RARα-PLZF in APL pathogenesis and to identify putative downstream target genes, gene expression profiling was undertaken using Affymetrix (Santa Clara, CA) U133 GeneChips (38) in primary APL blasts, comparing cases in which both reciprocal fusion products were expressed with those in which only PLZF-RARα was formed. Analysis of four cases with evaluable RNA revealed up-regulation of 111 genes (>2-fold, P < 0.05) in the presence of RARA-PLZF, including CRABPI [see supporting information (SI) Fig. 7 A and C; data have been deposited in the Gene Expression Omnibus (GEO) database (accession no. GSE8510)]. Real-time quantitative PCR showed 50-fold higher expression of CRABPI in APL with expression of RARA-PLZF compared with cases in which PLZF-RARA was the sole fusion transcript, and it showed 5-fold higher expression than in APL with the PML-RARA fusion (SI Fig. 7B). In accordance with these data, a very high-level expression of mCRABPI protein was observed in APL blasts derived from PLZF-RARα/RARα-PLZF double-transgenic mice when compared with PML-RARα blasts (Fig. 1A).

Fig. 1.

CRABPI is regulated by PLZF and RARα-PLZF through a remote intronic BS. (A) Western blot showing overexpression of CRABPI in PLZF-RARα/RARα-PLZF transgenic APL blasts compared with blasts from PML-RARα transgenic mice and KG1 cells. β-actin was used as a loading control. (B) Schematic representation of the genomic structure of the human CRABPI gene. Black boxes represent exons, and dark gray boxes represent intronic and promoter regions. In silico search of the human genomic CRABPI gene revealed only one putative PLZF DNA BS, which was located at the end of intron 3–4 (asterisk and underlined text). The location of primers used to amplify the region containing the BS are denoted in bold italic text.

Identification of a Consensus PLZF Binding Site Within the CRABPI Locus.

To investigate the possibility that RARα-PLZF and wild-type PLZF may directly regulate CRABPI expression, we first examined the CRABPI genomic sequence for potential PLZF binding sites (BSs). This approach revealed the presence of a single 9-bp motif (CATGTCATG) related to the PLZF BS previously described in the Hoxb2 promoter (29). Unexpectedly, this motif was not found in the promoter region but was located 5 kb downstream in the last intron (Fig. 1B). This putative regulatory element was well conserved among several mammalian species (SI Fig. 8).

PLZF Mediates Repression at the CRABPI Intronic Binding Site.

A short DNA fragment including the putative CRABPI PLZF BS (CRABPI-BS) was cloned in a luciferase reporter plasmid and used to assess the transcriptional activity of PLZF through this site. Full-length PLZF was able to repress the reporter when this fragment was present; as predicted, this activity was lost with a truncated version of PLZF lacking the N-terminal repressor domain (PLZF^ΔPOZ) but retaining DNA-binding capacity (SI Fig. 9A). In accordance with previous analysis of the Hoxb2 promoter (29), a transversion from T to G within the BS was able to abrogate the effect of PLZF on the plasmid reporter, whereas point mutation immediately outside the putative BS did not (SI Fig. 9B). Binding activity at this site was further assessed by ChIP assay, confirming that the former single point mutation was enough to abrogate PLZF binding. Acetylation of the surrounding DNA was assessed; this revealed a strong correlation between PLZF binding to the intronic sequence and complete deacetylation of the surrounding region (SI Fig. 9C). ChIP assays were carried out throughout the CRABPI genomic locus including the promoter region (−847 to +120) to assess the DNA-binding activity of PLZF. Specific binding of PLZF and RARα-PLZF to DNA was only observed in the intron 3–4 segment containing the PLZF BS (SI Fig. 9D).

PLZF and RARα-PLZF Exert Differential Activities at the PLZF Binding Site in the CRABPI Locus.

Transfection of increasing amounts of PLZF in the presence of the CRABPI-BS reporter vector led to decreasing transcriptional reporter activity (Fig. 2A). Conversely, transfection of increasing amounts of RARA-PLZF led to increased transcriptional activity. Cotransfection of RARA-PLZF and PLZF led to a net increase in transcriptional activity, although less than with RARA-PLZF alone. To investigate the mechanism of this effect of the chimera, we used a ChIP assay to assess recruitment of proteins at the DNA BS contained in the CRABPI-BS reporter vector. Antibodies to PLZF and RARα-PLZF (anti-PLZF and anti-Flag, respectively) were used to pull down specifically the two proteins and to investigate the proteins recruited in vivo at this site (Fig. 2B). As expected, no DNA was pulled down when the two proteins were not expressed, and the H3 histones in the surrounding region were acetylated. PLZF induced complete deacetylation of the region, as assessed by histone H3 status, as well as DNA methylation (SI Fig. 10), whereas no such effect was observed with RARα-PLZF. When the two proteins were coexpressed, the genomic region surrounding the CRABPI-BS lay within open chromatin with acetylated histones (Fig. 2B, anti-AcH3 in lane D compared with PLZF alone in lane C). RARα-PLZF seemed to compete for the DNA BS with PLZF (Fig. 2B; compare PLZF binding in lanes C and D, third row, with RARα-PLZF binding in lanes B and D, second row), leading to an increased acetylation level and transcriptional activity. To investigate this in the context of normal genomic DNA, we used a PLZF-expressing hematopoietic cell line KG1 and ectopically expressed RARA-PLZF. ChIP assay showed binding of PLZF protein to the endogenous genomic CRABPI-BS present in KG1 cells (Fig. 2C, lane 3) in the presence of transiently expressed PLZF-RARα^Flag or empty vector (data not shown). As expected, RARα-PLZF^Flag, but not PLZF-RARα^Flag, was able to bind to the CRABPI-BS (Fig. 2C, lanes 8 and 4, respectively). Furthermore, the RARα-PLZF^Flag chimeric protein was shown to compete with the wild type PLZF for the BS on genomic DNA (Fig. 2C, lane 7). Additionally, ChIP assays were performed in primary murine cells, and comparable PLZF binding activity and competition were observed at the mouse genomic DNA BS (data not shown).

Fig. 2.

Relative binding of PLZF and RARα-PLZF at the CRABPI intronic regulatory element. (A) CRABPI_{(intron3–4)-}tkLuc plasmid was used to assess the effects of PLZF (black square), RARα-PLZF (gray circle), and the two transcription factors (PLZF+RARα-PLZF, gray square) on the transcriptional activity of the reporter. Increasing amounts of each expression vector were cotransfected into 293T cells in the presence of the luciferase reporter (same amount for each point). Luciferase activity was determined 16 h later. (B) ChIP assay was performed in parallel with the luciferase assay to assess the acetylation level in the proximity of the PLZF BS (Anti-AcH3), together with the binding of the chimeric RARα-PLZF (Anti-Flag) and PLZF (Anti-PLZF) proteins. The level of histone H3 protein in the vicinity of this site (Anti-H3) was used as a positive control for the ChIP procedure and to normalize the quantity of input DNA at each transfection point. Lane A, CRABPI_{(intron3–4)-}tkLuc reporter alone; lane B, luciferase reporter and RARα-PLZF^Flag; lane C, luciferase reporter and PLZF; lane D, luciferase reporter, PLZF, and RARα-PLZF^Flag. (C) Endogenous PLZF DNA-binding activity in the hematopoietic KG1 cell line. (Left) ChIP assay shows binding of PLZF protein to the endogenous genomic CRABPI-BS present in KG1 cells (lanes 3 and 7, immunoprecipitation done with polyclonal anti-PLZF antibody) in the presence of transiently expressed PLZF-RARα^Flag or RARα-PLZF^Flag. Binding of the chimeric proteins was assessed using a monoclonal anti-Flag antibody (lanes 4 and 8). As expected, RARα-PLZF^Flag, but not PLZF-RARα^Flag, was able to bind to the CRABPI-BS. It appears that only the RARα-PLZF^Flag chimeric protein is competing with wild-type PLZF for the BS on genomic DNA. An irrelevant polyclonal antibody was used as a negative control in the ChIP procedure (lanes 1 and 5). (Right) Expression of the chimeric fusion proteins was confirmed by Western blotting.

Identification of Factors Recruited to the CRABPI Intronic Binding Site by Wild-Type PLZF and the RARα-PLZF-Associated Chimera.

To characterize the factors involved in the modulation of CRABPI gene activity initiated by PLZF and RARα-PLZF binding, we selected candidate target proteins for evaluation in re-ChIP assays (39). Briefly, ChIP was initially performed with an antibody recognizing either PLZF or RARα-PLZF^Flag to specifically purify the respective protein complexes. Then, a second immunoprecipitation (re-ChIP) assay was performed with specific antibodies raised against a number of transcriptional regulators by using these purified complexes as bait.

ChIP patterns obtained on chromatin prepared from transfected 293T cells demonstrated that a number of proteins are engaged at the PLZF CRABPI-BS (Fig. 3Left). Importantly, proteins present on the intronic CRABPI region in the presence of wild-type PLZF are associated with repressor functions, including N-CoR (lane 5), HDAC1 (lane 4), and DNMT1 (lane 7) (similar results were observed in untransfected KG1 myeloid cells, SI Fig. 11); however, methyl-binding proteins (40) were not detected (Fig. 3, see lanes 8 and 9). Conversely, RARα-PLZF chimeric protein did not recruit any members of the repressor complex to DNA, but was associated with recruitment of the histone acetyl-transferase (HAT) p300 (Fig. 3, lane 6, second row). Moreover, when PLZF and RARα-PLZF were coexpressed, the pattern of complex recruitment was transformed from a repressor to an activator profile (Fig. 3, third and fourth rows).

Fig. 3.

PLZF and RARα-PLZF recruit repressor and activator complexes, respectively. (Left) Chromatin prepared from transiently transfected 293T cells (PLZF and/or RARα-PLZF^Flag) were subjected to the ChIP procedure with anti-PLZF or anti-Flag antibodies, shown on the left, and were again immunoprecipitated using antibodies shown at the top of the image. (Right) Expression of RARα-PLZF and PLZF was confirmed by Western blotting.

PLZF Mediates Propagation of a Repressive Chromatin Environment to the CRABPI Promoter from the Downstream Binding Element.

To investigate the dynamics of repression of the CRABPI locus by PLZF, time-course studies were performed in 293T cells, which do not express significant levels of PLZF (SI Fig. 11). After transfection, as soon as PLZF protein levels were detectable by Western blot analysis, DNA binding at the CRABPI intronic BS was observed (Fig. 4A) and was accompanied by local deacetylation within a 510-bp region flanking the PLZF BS (Fig. 4B, Anti-acH3 in row iv). Over the course of the next 11 h, deacetylation spread progressively upstream and involved the promoter by 16 h, at which stage recruitment of MBD1 (Fig. 4B, lanes 16 and 24, anti-MBD1 in rows i and ii) was observed. Recruitment of MeCP2 was only observed in the promoter region at 24 h (data not shown). In contrast, binding of RARα-PLZF to the CRABPI BS did not induce any change in the acetylation status of the locus or recruitment of the methyl binding proteins (Fig. 4). In this cell line in which CRABPI protein is expressed, Western blot analysis did not show any augmentation in the presence of RARα-PLZF; conversely, expression of PLZF led to a significant down-regulation of CRABPI (Fig. 4A, fourth row).

Fig. 4.

PLZF propagates a repressive chromatin environment from a distant BS to silence the CRABPI promoter, which is overcome by RARα-PLZF. (A) Expression vectors for PLZF and RARα-PLZF (200 ng) were transfected into 293T cells. Levels of transiently expressed proteins were monitored at various time points (0, 2, 5, 12, and 24 h) by Western blot analysis using an anti-Flag (Anti-Flag Western) antibody, and their binding activities were monitored by ChIP using an anti-Flag antibody (Anti-Flag ChIP) at the same time points. Levels of expressed CRABPI protein throughout the time-course experiment were assessed by Western blotting using an anti-CRABPI antibody (Anti-CRABPI Western). The anti-Flag antibody was used to pull down the Flag-tagged proteins attached to the genomic DNA of the 293T cells and was tested to determine whether the overexpressed proteins were able to bind the endogenous genomic CRABPI sequence. (B) The effect of PLZF and its chimeric protein RARα-PLZF on the acetylation levels of histone H3 and the recruitment of MBD1 were assessed throughout the length of the CRABPI gene. Various segments of genomic DNA situated in the intronic regions 3–4 and 2–3, as well as in the promoter region of the CRABPI gene, were tested at various time points after PLZF and RARα-PLZF transfection into 293T cells for their respective levels of histone H3 acetylation and the abundance of MBD1 proteins interacting with DNA (Anti-AcH3 and Anti-MBD1, respectively). Results are shown as gel pictures and as quantification bar graphs (gray bars, histone H3 acetylation; black bars, MBD1 recruitment to DNA). All ChIP experiments were performed from a single chromatin preparation for each time point. Throughout the ChIP experiment, efficiency of the precipitation was assessed by using an anti-histone H3 (Anti-H3) antibody as a positive control.

PLZF Induces DNA Methylation of the CRABPI Promoter.

We next considered the consequences of recruitment of DNMT1, MBD1, and MeCP2 induced by the binding of PLZF at the distant downstream BS on the methylation status of the CRABPI promoter by using restriction digest of genomic DNA with methylation-sensitive and -insensitive enzymes (AvaI and AvaII, respectively) followed by PCR (Fig. 5A and SI Materials and Methods). The promoter was unmethylated in 293T cells, which do not express PLZF at a detectable level (Fig. 5B and SI Fig. 11), and confirmed to be methylated in hematopoietic KG1 (Fig. 5C) and HL60 (Fig. 6C) cells, which express PLZF (ref. 41 and SI Fig. 11). Expression of full-length PLZF in 293T cells led to induction of DNA methylation at the CRABPI promoter within 24 h (Fig. 5B), whereas expression of RARα-PLZF in KG1 and HL60 cells led to demethylation of the promoter (Figs. 5C and 6C) and up-regulation of CRABPI by Western blot analysis (Figs. 5D and 6D). No such effect was observed when PLZF-RARα or empty vectors were expressed. The impact of RARα-PLZF expression on the methylation status of the CRABPI promoter was also determined in primary human PLZF-RARA+ APL blasts. The promoter was methylated in cases in which PLZF-RARA was the sole fusion transcript expressed and was hypomethylated where RARA-PLZF was coexpressed (Fig. 5E), in accordance with the cell line data (Fig. 5C).

Fig. 5.

Methylation status of the CRABPI promoter correlates with RARα-PLZF expression in primary APL cells. Methylation status was determined by digesting the genomic DNA from cell lines and patient samples by AvaI and AvaII endonucleases. These two enzymes behave differently in the presence of methylated DNA, being sensitive and insensitive, respectively. (A) Schematic representation of the genomic structure of the human CRABPI gene. Black boxes represent exons; dark gray boxes represent intronic and promoter regions. The CpG islands located at the beginning of the CRABPI gene are indicated by a light gray box, and AvaI and AvaII endonuclease sites are indicated by arrowheads and circles, respectively. (B) Effect of overexpression of PLZF on the methylation status of the CRABPI promoter. Ectopic expression of PLZF in 293T cells induced methylation in this region, as indicated by the appearance of a specific band in lane avaI at 24 h after transfection, reflecting the inability of the AvaI endonuclease to cut the genomic DNA at this specific location. Lane −, no enzyme input control; lane avaI, AvaI digest; lane avaII, AvaII digest (as a control for digestion). (C) Effect of overexpression of RARα-PLZF and PLZF-RARα in the PLZF expressing the KG1 cell line. In the untransfected, empty vector and PLZF-RARα-transfected KG1 cells, the CRABPI promoter is found to be methylated because of the endogenously expressed PLZF (lane −, untransfected). Interestingly, only overexpression of the chimeric RARα-PLZF protein induced specific demethylation of the promoter (lane avaI, RARα-PLZF). (D) Western blot detection of the CRABPI and -II proteins in KG1 cells. PLZF-RARα^Flag and RARα-PLZF^Flag were overexpressed for 24 h, and levels of protein expression were assessed. Up-regulation of CRABPI was only observed in the KG1 cells expressing the RARα-PLZF^Flag chimeric protein (lane RARα-PLZF) relative to levels of CRABPII and β-actin, which remained unchanged. (E) Methylation status of the CRABPI promoter in primary PLZF-RARA+ leukemic blasts from APL patients, according to expression status of the reciprocal RARA-PLZF fusion transcript. As shown, cases 1 and 2 expressing the reciprocal RARA-PLZF had low levels of DNA methylation compared with cases in which PLZF-RARA was the sole transcript expressed (cases 4 and 5; all avaI lanes).

Fig. 6.

RARα-PLZF confers retinoid resistance in a CRABPI-dependent fashion. (A and B) RARα-PLZF blocks RA-induced differentiation of HL60 cells in a CRABPI-dependent fashion. HL60 cells were plated in selection medium with or without RA at 10⁻⁶ M concentration and were treated with antisense (CrI AS) and missense (CrI MS) CRABPI oligonucleotides for 7–10 days; the number of CFU-G colonies formed was determined after 7–10 days of culture. In the presence of culture medium alone, HL60 cells spontaneously produce CFU-G (A1 and B). RA led to a marked increase in CFU-G (A2 and B), which was not influenced by CRABPI antisense or missense oligonucleotides. (A and D) Expression of RARα-PLZF blocked RA-induced differentiation (A6), which was restored (A7) after knockdown of CRABPI by the presence of the antisense oligonucleotide (D), but was unaffected by the missense oligonucleotide (A8 and D). (C) Methylation status of CRABPI promoter in HL60 cells according to expression of RARα-PLZF. The CRABPI promoter is methylated in wild-type HL60 cells but is demethylated in the presence of RARα-PLZF. CRABPI antisense (CrI AS) was confirmed to have no impact on the methylation status of the CRABPI locus. (D) (Left) Western blot showing that expression of RARα-PLZF (detected by anti-Flag antibody) leads to induction of CRABPI protein levels that were effectively knocked down by CRABPI antisense (CrI AS). β-Actin was used as a loading control. (Right) Corresponding levels of CRABPI expression normalized to the ABL housekeeping gene.

Induction of CRABPI Due to RARα-PLZF Expression Mediates Retinoid Resistance.

To investigate the impact on sensitivity to RA, RARα-PLZF was expressed in myeloid HL60 cells, which normally undergo terminal differentiation in response to this agent (42). Clonogenic progenitor assays were used in which differentiation leads to the formation of granulocytic colonies [(CFU-granulocyte (CFU-G)] (43). RARα-PLZF expression led to up-regulation of CRABPI and impaired RA-dependent differentiation (Fig. 6). Retinoid sensitivity was restored in the presence of RARα-PLZF by knockdown of CRABPI using an antisense oligonucleotide (44), but not with the corresponding missense oligonucleotide (Fig. 6 A and B). Together these data show that RARα-PLZF confers retinoid resistance that is specifically mediated by up-regulation of CRABPI.

Discussion

PLZF was first identified through its involvement in APL with the t(11;17) translocation (45), and recent studies have shown that it plays an important role in stem cell maintenance and embryonic development (46, 47). PLZF binds DNA in a sequence-specific manner, leading to repression of genes containing PLZF BSs within their promoter regions (27–29, 48, 49). It has also been reported that PLZF can cooperate with other POZ family members, including the related BCL-6 protein that is associated with diffuse large-cell lymphoma (50, 51). Both PLZF and BCL-6 are potent transcriptional repressors (18, 19, 21, 52, 53) and can interact, cooperating in a mechanism of gene silencing in hematopoietic tissue (54). Studies to date have indicated that PLZF mediates repression at two levels. In the case of cyclin A2 (27), c-myc (28), Hoxb2 (29), Tpo receptor (55), and VLA-4 antigen (56), PLZF BSs have been documented in the promoter regions, leading to gene silencing at a local level. Whereas, the Hoxd11 gene is regulated by an alternative mechanism involving long-range repression associated with DNA looping and bridging of distant PLZF BSs (48). A similar process may also involve the Polycomb protein family member, Bmi-1 (48), which can mediate long-distance repression of target promoters (57) and which is an interaction partner of PLZF.

A number of previous studies have shown that PLZF-mediated repression involves recruitment of corepressors and histone deacetylation of target promoters (28, 29, 55, 56). Here, we show that PLZF is also involved in DNA methylation through recruitment of DNMT1 and can lead to gene silencing through an alternative mechanism involving propagation of a repressive chromatin environment after sequence-specific binding to remote PLZF consensus sites. We define such a site, located within an intron, 5 kb downstream from the promoter of the CRABPI gene, which is implicated in regulation of retinoid signaling. Through time-course experiments in conjunction with ChIP assays, PLZF binding to the CRABPI intronic region was initially found to lead to recruitment of HDAC1 and DNMT1 with localized deacetylation and DNA methylation. Histone deacetylation, accompanied by a wave of methylation, then advanced toward the promoter, reaching the CpG islands where MBD1 and MeCP2 were recruited, associated with gene silencing. Interestingly, expression of the t(11:17) reciprocal fusion gene product RARα-PLZF, but not PLZF-RARα, competed with PLZF at the CRABPI intronic BS, overcoming PLZF-mediated repression.

In contrast to PLZF, which recruited the N-CoR/HDAC complex to the CRABPI intronic BS, RARα-PLZF recruited an alternative p300-associated complex, inducing up-regulation of CRABPI (SI Fig. 12). The recruitment of p300 by RARα-PLZF is not unexpected given that wild-type PLZF is able to interact with p300 protein via the zinc finger region, which is retained in this fusion product (58). This activator complex recruited by RARα-PLZF induced hypomethylation of the CRABPI promoter, which was also observed in primary APL samples in an RARα-PLZF-dependent manner. This target is of particular interest, given that CRABPI is well established as a mediator of retinoid resistance in a number of cellular systems (34–37). Moreover, CRABPI has been implicated as a mediator of secondary clinical resistance to RA in APL with the PML-RARα fusion (59), strongly suggesting that, in the context of APL cells, its expression decreases RA-sensitivity. Its induction by RARα-PLZF blocked the terminal differentiation of myeloid leukemia cells that are normally sensitive to RA, demonstrating that up-regulation of CRABPI mediates primary resistance to RA in human leukemia. Finally, the in vivo relevance of these observations was obtained from the comparison of PML-RARα and compound PLZF-RARα+RARα-PLZF APLs, where in similar leukemic promyelocytes, the levels of CRABPI protein were found to be very different (Fig. 1A). Therefore, up-regulation of CRABPI mediated by RARα-PLZF, acting in concert with PLZF-RARα, could contribute to the differentiation block, as well as to the retinoid insensitivity that characterizes this subset of leukemia (SI Fig. 12).

Although chimeric fusion proteins generated by chromosomal translocations commonly exhibit corepressor recruitment capability, mediating repression of genes implicated in hematopoietic differentiation, we show in this study that the reciprocal RARα-PLZF fusion protein is functionally distinct, recruiting coactivator complexes and potentially overcoming the repression of PLZF target genes including CRABPI (SI Fig. 12). In some respects, this is analogous to the observation that the AF10 component of the MLL-AF10 fusion generated by the t(10;11)(p12;q23) translocation in AML recruits hDOT1L, involved in histone methylation associated with transcriptional activation, leading to up-regulation of HoxA9 contributing to leukemogenesis (60). These studies underline the contribution of the fusion partner in the pathogenesis of AML. Moreover, the present study reveals the capacity for reciprocal fusion gene products to mediate distinct but complementary epigenetic effects underlying the leukemic phenotype.

Materials and Methods

ChIP and Re-ChIP Assays.

293T cells in 10-cm plates were transfected with 5 μg of PLZF-RARA or RARA-PLZF expression vectors by using the calcium phosphate precipitation method. Immunoprecipitation of plasmid DNA plus associated histones was carried out at various times after transfection according to a previously published protocol (58, 61), with some modifications (detailed in SI Materials and Methods). In re-ChIP experiments (62), complexes were eluted by incubation for 30 min at 37°C in 25 μl of 10 mM DTT and were subjected to the ChIP procedure. DNA sequences were detected by semiquantitative PCR (see SI Materials and Methods for primer sequences).

Clonogenic Progenitor Assay.

HL60 cells were cotransfected by electroporation with empty or RARα-PLZF pSG5 expression vector (Stratagene, La Jolla, CA) and the pMMAMNeo selection vector (Clontech, Mountain View, CA). After transfection, cells were cultured in complemented RPMI in G418 at 600 ng/ml for 1 day and were plated at 2 × 10⁵ cells per ml in 1 ml of mixture containing MethoStem medium with 15% FCS, 1% antibiotics, G418 (600 ng/ml) 0.3% agar (PAA Laboratories, Linz, Austria), ligand, and phosphorothioate oligonucleotides. The culture plates were incubated at 37°C and in 5% CO₂ atmosphere. At 7–10 days, granulocytic colonies (CFU-G) were scored with an inverted microscope.