|Home | About | Journals | Submit | Contact Us | Français|
PML-RAR is an oncogenic transcription factor forming in acute promyelocytic leukemias (APL) because of a chromosomal translocation. Without its ligand, retinoic acid (RA), PML-RAR functions as a constitutive transcriptional repressor, abnormally associating with the corepressor-histone deacetylase complex and blocking hematopoietic differentiation. In the presence of pharmacological concentrations of RA, PML-RAR activates transcription and stimulates differentiation. Even though it has been suggested that chromatin alteration is important for APL onset, the PML-RAR effect on chromatin of target promoters has not been investigated. Taking advantage of the Xenopus oocyte system, we compared the wild-type transcription factor RARα with PML-RAR as both transcriptional regulators and chromatin structure modifiers. Without RA, we found that PML-RAR is a more potent transcriptional repressor that does not require the cofactor RXR and produces a closed chromatin configuration. Surprisingly, repression by PML-RAR occurs through a further pathway that is independent of nucleosome deposition and histone deacetylation. In the presence of RA, PML-RAR is a less efficient transcriptional activator that is unable to modify the DNA nucleoprotein structure. We propose that PML-RAR, aside from its ability to recruit aberrant quantities of histone deacetylase complexes, has acquired additional repressive mechanisms and lost important activating functions; the comprehension of these mechanisms might reveal novel targets for antileukemic intervention.
Acute promyelocytic leukemia (APL), a subset of acute myeloid leukemias, is morphologically characterized by the accumulation of malignant myeloid cells blocked at the promyelocytic stage of maturation. Genetically, it is characterized by a translocation that fuses the retinoic acid (RA) receptor α (RARα) gene on chromosome 17 with either the promyelocytic leukemia (PML) gene on chromosome 15 or the PML zinc finger (PLZF) gene on chromosome 11 (11). The resulting RAR chimeric genes encode fusion proteins PML-RAR and PLZF-RAR. Ectopic expression of RAR fusion proteins in hemopoietic precursor cells blocks their ability to undergo terminal differentiation, while transgenic mice overexpressing PML-RAR or PLZF-RAR develop an APL-like phenotype, indicating the causal role of these fusion proteins in APL (4, 9, 12). Structure-function analysis of PML-RAR revealed that the RAR component of the fusion protein is indispensable for its ability to impair terminal differentiation (12).
RARα is a member of the superfamily of nuclear hormone receptors that bind to specific RA response elements as heterodimers with retinoid X receptor (RXR) (21). The activity of RAR/RXR is influenced by its physiological ligand, RA. In the absence of RA, the heterodimer recruits the nuclear corepressor N-CoR/histone deacetylase complex (HDAC), suggesting that it silences its target genes by altering chromatin structure (23, 30). Upon hormone addition, the corepressor complex is released and the receptor associates with transcriptional coactivators. A rapidly expanding repertoire of coactivators has recently been isolated, including proteins possessing histone acetyltransferase (HAT) activity (18). The observation that transcriptional activation is associated with recruitment of HAT has reinforced the model in which modification of chromatin structure targeted by RAR/RXR contributes to transcriptional control. However, the precise molecular mechanisms that bring the RA signal from the nuclear receptors to the transcriptional machinery have yet to be elucidated.
It has recently been demonstrated that transcriptional activation by RAR/RXR requires both ATP-driven chromatin remodeling and coactivator HAT activities and that these multiprotein complexes act in a temporally ordered manner (7). A simplistic model for the coactivator recruitment by nuclear receptors involves different steps. First, proteins that remodel nucleosomal structures are recruited to the liganded receptors. Subsequently, TRAP/DRIP complexes can replace the previous ones and bind the receptors. The recruitment of RNA polymerase II (PolII) to these coactivators completes this second step in transactivation (10, 26).
In divergence from this sequential model, it has recently been proposed that the cofactors essential for hormone-dependent activation are not all recruited through their direct interactions with the nuclear receptors but can be recruited through a combinatorial effect of direct or indirect interactions (15). However, these two models are not considered mutually exclusive and might be receptor specific.
PML-RAR retains both DNA binding domains and ligand binding domains of RARα. It shows an affinity for RA comparable to that of wild-type RARα (2, 8) and binds to RA response elements as either homodimers or multimeric complexes containing or not containing RXR (20). Since its ability to block hematopoietic differentiation depends on an intact DNA binding domain, RA target genes are thought to represent downstream effectors of PML-RAR (11). Mechanistically, the fusion protein is thought to block differentiation by constitutively silencing RA-responsive genes involved in the control of differentiation of hematopoietic precursor cells. In accordance, it has been demonstrated that the fusion protein forms stable complexes with N-CoR/HDAC in the absence of RA or in the presence of physiological concentrations of RA, while pharmacological doses of ligand provoke the release of HDAC (13, 14, 19). This abnormal recruitment of N-CoR by PML-RAR derives from the coiled-coil region of PML, which permits the formation of PML-RAR oligomers in vivo (20, 24). It is generally accepted that APL transformation correlates with PML-RAR repressive functions on the basis of the following observations: (i) mutations of the PML-RAR N-CoR binding site(s) impair the biological activity of the fusion protein, (ii) oligomerization per se is sufficient to activate the oncogenic potential of the transcription factor RAR, (iii) large doses of RA induce disease remission in PML-RAR APL patients (11), and (iv) HDAC inhibitors combined with RA enhance terminal differentiation of APL cells. However, studies of the mechanisms by which PML-RAR regulates transcription are limited by the fact that repression has not been analyzed in a native chromatin context. Xenopus oocytes are an attractive in vivo system with which to perform these studies since microinjection of double-stranded DNA leads to progressive chromatinization of the template (17). Moreover, this system, unlike mammalian cells, contains undetectable levels of endogenous receptors, thus allowing unambiguous evaluation of the properties of the main regulators of APL.
In this work, we used Xenopus oocytes to perform a comparative analysis of the wild-type transcription factor RARα and PML-RAR as both transcriptional regulators and modifiers of chromatin structure. As a target promoter, we used the RA receptor β2 (RARβ2) promoter, a well-defined in vivo target of PML-RAR (6). Our results demonstrate that unliganded PML-RAR is a different repressor compared to the wild-type protein. In fact, whereas a heterodimer of RAR/RXR is required to silence RARβ2, the oncoprotein is a powerful repressor in the absence of its partner. Surprisingly, only PML-RAR owns a repressive pathway independent of nucleosome deposition and histone deacetylation. Also, in the presence of RA, the two proteins behave differently; in fact, RAR/RXR is a significantly more powerful activator than PML-RAR. We discuss possible molecular mechanisms justifying the observed phenomena.
Chromatin assays confirmed the new properties of PML-RAR. In fact, in the absence of RA, RAR/RXR binding to DNA does not produce topological changes, whereas unliganded PML-RAR closes the chromatin configuration. Since trichostatin A (TSA) can overcome this effect, we can state that the aberrant recruitment of HDAC activities is responsible for this closing effect. Analysis of DNase I-hypersensitive sites revealed that wild-type heterodimer and PML-RAR bind to minichromosomes, but only liganded RAR/RXR modifies the nucleoprotein structure organized on DNA. The absence of chromatin reconfiguration in the presence of RA-bound PML-RAR confirms that the fusion protein has lost some activating function(s).
To generate the RARβ2pCAT plasmid, a PCR fragment of 138 bp containing the murine RARβ2 promoter (−124 to +14) was subcloned into the SphI-AccI sites of the pCAT vector. The Gal4RARβ2pCAT plasmid was obtained by inserting a PCR fragment of 146 bp, containing five Gal4-binding sites, amplified from pG5-HSPCAT (17) into the HindIII site of the RARβ2pCAT plasmid. The expression vectors for RAR, RXR, PML-RAR, and GAL4VP16 have been previously described (17, 22). All of the constructs prepared by inserting a PCR amplification product were confirmed by DNA sequencing.
The preparation of Xenopus stage VI oocytes and the microinjection procedure were as previously described (16). For the expression of exogenous transcription factors, full-length capped RNAs were synthesized by SP6 (RXR, PML/R) or T7 (RAR) RNA polymerase transcription of linearized templates with an mMessage mMachine kit (Ambion). Thirty-two nanoliters of in vitro-synthesized mRNA was injected into the oocyte cytoplasm. Five nanograms of PML-RAR, 3 ng of RAR, and 1.5 ng of RXR mRNAs were injected into the cytoplasm, and 16 h later, the oocytes were injected, in the nucleus, with 32 nl of double-stranded RARβ2pCAT or GAL4RARβ2pCAT (45 ng/μl) and incubated for 16 h or for the lengths of time stated in the figure legends, in the absence or presence of RA (10 μM), TSA (330 nM), or RA and TSA. Healthy oocytes were collected and assayed for transcript levels, chromatin assembly, DNA recovery, and protein expression. To express Gal4VP16, 32 nl of in vitro full-length capped mRNA (200 ng/μl), synthesized by SP6 RNA polymerase, was injected into the oocyte cytoplasm or 0.5 to 2.5 ng of recombinant protein was injected into the nucleus.
Twenty oocytes were collected and homogenized in 200 μl of 0.25 M Tris-HCl (pH 8.0). From half of the sample, RNA was extracted with Eurozol (EuroClone) and an amount equivalent to six oocytes was analyzed by primer extension as previously described (16).
The 30-mer 5′-GGTGGTATATCCAGTGATTTTTTTCTCCAT-3′, which is complementary to the cat gene, was used as a primer. The extension product was resolved on a 6% sequencing gel and visualized by autoradiography. The transcription signal was quantified with a PhosphorImager (Molecular Dynamics).
For DNA analysis, half of the homogenate was incubated for 2 h at 37°C in 30 mM EDTA-20 mM Tris-HCl (pH 7.5)-1% sodium dodecyl sulfate (SDS)-500 μg of proteinase K per ml. DNA was extracted twice with phenol-chloroform and ethanol precipitated. After RNase A treatment, samples corresponding to 1.5 oocytes were subjected to 1% agarose gel electrophoresis, transferred on a Hybond N membrane (Amersham), and hybridized against a randomly primed pCAT plasmid with PerfectHyb Plus hybridization buffer (Sigma).
Thirty oocytes were homogenized at 25 μl per oocyte in protein extraction buffer (70 mM KCl, 20 mM HEPES [pH 7.5], 1 mM dithiothreitol, 5% sucrose) with 3 mM CaCl2. The homogenate was incubated at 20°C in the presence of micrococcal nuclease (MNase; 4 U per oocyte; Worthington) for 4, 10, and 35 min. The reaction was stopped by addition of EDTA, Tris-HCl (pH 7.5), and SDS to final concentrations of 30 mM, 20 mM, and 1%; the mixture was then incubated for 1 h at 37°C with 500 μg of proteinase K per ml.
DNA was purified as described by Landsberger and Wolffe (17), resolved on a 1.5% agarose gel, and hybridized with the probes described in the figure legends.
Injected oocytes (25 per sample) were homogenized in protein extraction buffer (18 μl per oocyte) with 5 mM MgCl2. The homogenate was divided into three fractions and digested at 20°C for 2.5 min with 3, 10, and 15 U of DNase I (Worthington), respectively. The reaction was stopped by addition of an equal volume of 20 mM Tris-HCl (pH 7.5)-30 mM EDTA-1% SDS-500 μg of proteinase K per ml. After the purification procedure (17), each DNA sample was resuspended in 100 μl of 1× NcoI buffer, digested with 10 U of NcoI for 2 h at 37°C, precipitated, and finally resolved on a 1.5% agarose gel. Southern blotting was performed with an NcoI-EcoRI fragment from the RARβ2CAT vector, adjacent to the promoter sequence.
DNA from 5 to 10 injected oocytes was extracted and purified as described above. DNA equivalent to 1.5 oocytes was separated on a 1% agarose gel in 1× Tris-borate-EDTA containing 10 μg of freshly prepared chloroquine per ml in both the gel and the running buffer at 50 V for 22 h in the dark, blotted to nylon membrane, and hybridized with a randomly primed, 32P-labeled pCAT vector. In the two-dimensional supercoiling assay, DNA was first separated on 1% agarose gel containing 10 μg of chloroquine per ml at 50 V for 22 h. The second dimension was run in the presence of 15 μg of chloroquine per ml at 50 V for 20 h.
To investigate the behavior of PML-RAR, RAR, and RXR in a chromatin context, we used Xenopus oocytes and measured their transcriptional effects on the RARβ2 promoter (3). Briefly, oocytes were injected with mRNAs encoding nuclear receptors or left as uninjected controls and, after 16 h, injected with RARβ2 plasmid DNA. Injected oocytes were incubated for a further 16 h in the absence or presence of RA, allowing full assembly into chromatin (27) (schematized in Fig. Fig.1A).1A). Promoter activity was evaluated by primer extension and quantified with a PhosphorImager (Fig. (Fig.1B).1B). Moreover, by Southern blotting (Fig. (Fig.1C)1C) and Western blotting (Fig. (Fig.1D),1D), we verified the quality of the injections, as well as the presence in the injected oocytes of comparable amounts of DNA and of exogenous factors.
In control oocytes, levels of basal transcription from the RARβ2 promoter were low, and only slightly increased by RA, in accordance with the very low abundance of nuclear receptors in oocytes (Fig. (Fig.1B,1B, lanes 1 and 2) (22, 28). Exogenous expression of RAR or RXR per se did not influence basal RARβ2 transcription (lanes 3 and 5) but permitted ligand-dependent promoter activation, even though the RAR response was more pronounced (lanes 4 and 6). Simultaneous expression of RAR and RXR, instead, induced complete silencing of the promoter (Fig. (Fig.1B,1B, lane 7). Notably, PML-RAR repressed RARβ2 transcription regardless of the concomitant expression of exogenous RXR (Fig. (Fig.1B,1B, lanes 9 and 11), suggesting that PML-RAR does not require RXR for its repressive function. In the presence of RA, PML-RAR (alone or with RXR) activated transcription of the gene for RARβ2, although transactivation was significantly lower than that obtained with exogenous RAR or RAR plus RXR (Fig. (Fig.1B,1B, compare lane 10 or 12 with lane 4 or 8). Notably, this was not due to insufficient uptake of ligand into the oocytes, since it was not rescued by increasing RA concentrations (data not shown). To better evaluate the transcriptional repression by RAR-RXR and PML-RAR, we used a modified RARβ2 promoter containing five binding sites for the GAL4VP16 transcriptional activator (Fig. (Fig.1E).1E). Expression of GAL4VP16 led to a marked (sixfold) activation of transcription from this construct (Fig. (Fig.1E,1E, lanes 1 and 2). Concomitant expression of RAR-RXR or PML-RAR antagonized GAL4VP16 transactivation by about 40 and >90%, respectively (Fig. (Fig.1E,1E, lanes 3 and 4). Together, these results indicate that PML-RAR possesses altered transcriptional properties with respect to the wild-type receptor. In particular, in the unliganded state, it functions as a stronger repressor that is able to impose its silencing effect even in the absence of RXR; in contrast, PML-RAR possesses reduced ligand-dependent transactivating activity with respect to the wild-type receptor.
We next investigated the contribution of histone deacetylation and chromatin assembly to the transcriptional activities of PML-RAR. For this purpose, we evaluated the effects of the histone deacetylase inhibitor TSA, and we took advantage of the gradual chromatinization of injected DNA to examine whether a chromatin template is required for the observed regulation (27). On fully assembled minichromosomes (Fig. (Fig.2A),2A), TSA completely reversed the repressive effect of unliganded RAR/RXR, restoring transcription to the same levels obtained in TSA-treated control oocytes. In the presence of PML/RAR expression instead, TSA was unable to counteract basal transcriptional repression, suggesting that the fusion protein acquired a further repressive mechanism, independent of deacetylation. When given simultaneously with RA, TSA further increased gene expression in both RAR/RXR- and PML-RAR-expressing oocytes. As previously observed (Fig. (Fig.1B),1B), PML/RAR showed reduced ligand-dependent transactivating properties.
To investigate whether PML-RAR is able to inhibit gene expression through mechanisms independent of chromatin assembly, we analyzed the efficiency of transcriptional repression directed by the unliganded fusion protein as chromatin assembly proceeds (Fig. (Fig.2B).2B). As expected, wild-type RAR-RXR induced significant repression of activated (by GAL4VP16) RARβ2 transcription only after 16 h, when chromatin is formed. In contrast, unliganded PML-RAR inhibited gene expression at the earliest time point (2 h), and only a modest effect was observed with further chromatin assembly (16 h).
Finally, we analyzed the transactivating properties of receptors at 2 h in the absence of nucleosomes (Fig. (Fig.2C).2C). Addition of RA to RAR/RXR-expressing oocytes stimulated promoter activity more than 30-fold, revealing a significantly greater effect than in the presence of chromatin (compare Fig. 2A and C). This difference in transcriptional stimulation might be due to the existence in Xenopus oocytes, as in mammalian cells, of a phenomenon of attenuation of hormone-induced gene activation (5) and/or to a chromatin-mediated constraint of the transcriptional machinery recruitment. RARβ2 induction in the presence of PML-RAR and RA was dramatically inferior, confirming that in oocytes, the activating capabilities of this protein are severely compromised. In all of the samples, addition of TSA did not lead to any significant transcriptional activation, testifying to the absence of a well-defined nucleosomal array on the injected DNAs.
In conclusion, it appears that the wild-type heterodimer requires chromatin assembly to inhibit transcription and that this effect depends on HDAC activity. The repressive activity of PML-RAR, instead, is characterized by a further mechanism, which is independent of nucleosome deposition. Finally, RAR fusion to PML results in a marked reduction of its transactivating properties.
Structure-function analysis of PML-RAR demonstrated that abnormal recruitment of HDAC is crucial for its transforming ability (20, 24), suggesting that modifications of chromatin structure at target promoters represent an important mechanism of leukemogenesis. However, the effects of PML-RAR on the chromatin architecture of target genes remain uncharacterized. Therefore, we next compared the effects of RAR and PML-RAR on chromatin structure. Our first approach made use of an assay (supercoiling assay) based on the facts that each nucleosome constrains one negative superhelical turn and that modifications of the overall chromatin architecture can lead to changes in DNA topology.
Unliganded RAR or RAR/RXR did not lead to topological changes in DNA (Fig. (Fig.3A,3A, compare lane 3 with lanes 5 and 7). In contrast, PML-RAR, regardless the presence of RXR, induced a marked topological modification (lanes 9 and 11). To confirm that this effect was due to a net increase in negative superhelicity, we performed two-dimensional agarose gel electrophoresis. In the presence of PML-RAR, minichromosomes reached a more condensed conformation, gaining an average of two additional negative superhelical turns, compared to RAR-RXR (Fig. (Fig.3B3B).
Addition of RA to RAR/RXR-containing oocytes induced chromatin remodeling, imposing a configuration consistent with the loss of two nucleosomes (Fig. (Fig.3A,3A, lane 8). Remodeling strictly depended on the presence of RXR, since no changes in DNA superhelicity were obtained with liganded RAR (Fig. (Fig.3A,3A, compare lanes 5 and 6). PML-RAR, instead, responded to hormone addition even in the absence of RXR, with a reduction in negative superhelicity of 1 U (Fig. (Fig.3A,3A, lane 10). Addition of RXR, however, caused a bigger transition in chromatin structure, corresponding to the loss of two negative turns (Fig. (Fig.3A,3A, lane 12). Notably, the minichromosomes assembled in the presence of the oncoprotein were still characterized by a more condensed conformation (Fig. 3A and C).
We next analyzed the effects of TSA on DNA topology (Fig. (Fig.3D).3D). As previously reported (15, 28), no detectable changes occurred in minichromosomes assembled in control oocytes (Fig. (Fig.3D,3D, lane 4). In the presence of RAR/RXR, TSA treatment led to the loss of almost two histone octamers (compare lanes 6 and 7). The effect of TSA was identical to that of RA (lane 8). We then repeated the same assay with PML-RAR (lanes 9 to 11). TSA (lane 10) induced the loss of an average of three to four negative superhelical turns, allowing a chromatin conformation equivalent to that assembled in the presence of RAR/RXR and TSA (compare lane 10 with lane 7). Addition of RA (lane 11) had a more modest effect than the deacetylase inhibitor and did not increase the effect of TSA when given simultaneously (data not shown). No modifications of the described behavior were observed when RXR was expressed together with the fusion protein (data not shown). These results indicate that PML-RAR, in contrast to RAR-RXR, induces a remodeling of the general chromatin structure, imposing a more condensed configuration, and that this effect is relieved by the HDAC inhibitor.
We next investigated whether, in the presence of a strong transcriptional activator, the fusion protein maintains its chromatin-remodeling effect. As in previous experiments, the G5-RARβ2 DNA was microinjected after nuclear-receptor translation (Fig. (Fig.4A).4A). Two different amounts (0.5 and 2.5 ng) of the GAL4VP16 activator were added after chromatin assembly, oocytes were incubated for a further 6 h, and transcription was analyzed by primer extension (Fig. (Fig.4B)4B) and chromatin structure was analyzed by the supercoiling assay (Fig. (Fig.4C).4C). The G5-RARβ2 promoter was activated by the late addition of increasing amounts of GAL4VP16 (Fig. (Fig.4B).4B). Addition of RAR/RXR and, to a greater extent, PML-RAR reduced the GAL4VP16 transactivating effect (Fig. (Fig.4B,4B, compare lanes 6 and 7 with lanes 4 and 5). Analysis of the overall topology demonstrated that small amounts of GAL4VP16 did not lead to significant topological changes (Fig. (Fig.4C,4C, compare lanes 1 and 2), whereas greater amounts reduced by 2 to 3 U the number of nucleosomes assembled on the template (lane 3). This is consistent with previous observations suggesting that GAL4VP16 participates in nucleosome disruption through mechanisms other than DNA binding alone (17). Unexpectedly, simultaneous expression of high concentrations of GAL4VP16 and RAR/RXR further reduced the negative superhelical turns of the injected templates (compare lanes 3 and 5), suggesting that chromatin-bound RAR/RXR opens the chromatin configuration even though it inhibits gene expression. In the presence of PML-RAR and either low or high concentrations of GAL4VP16, the topoisomer distribution revealed a consistently more closed conformation, demonstrating the dominant ability of the fusion protein to make the chromatin structure compact (compare lanes 6 and 7 with lanes 2 to 5).
In summary, our data indicate that the unliganded wild-type heterodimer does not produce detectable chromatin remodeling while PML-RAR alters the overall chromatin structure, leading to a more closed conformation. This effect is maintained even in the presence of a strong transcriptional activator. Hormone-bound RAR/RXR targets a chromatin-remodeling activity, documented by the loss of two negative superhelical turns; similarly, the activated TR/RXR heterodimer decreases the DNA topology by two to three supercoils per binding site (28). In this respect, PML-RAR shows similar behavior, with the notable difference that it does not require RXR.
We further investigated chromatin remodeling induced by wild-type and oncogenic receptors through nuclease mapping of injected templates. Oocytes were microinjected with RARβ2 reporter DNA after nuclear-receptor expression, incubated with or without RA for 16 h, homogenized, and digested with increasing amounts of DNase I. Digestion products were analyzed by indirect end labeling (Fig. (Fig.5).5). In control oocytes, few DNase I-hypersensitive sites are present in the vector sequences whereas no fragments appear in the region including the RARβ2 regulatory sequences that go from 500 to 700 bp from the NcoI linearization site (Fig. (Fig.5A,5A, lanes 1 to 6; see also the diagram and the open box representing the promoter in the left part of Fig. Fig.5A).5A). This suggests that under conditions of low transcriptional activity, components of the basal transcriptional machinery are not stably associated with the promoter. Expression of unliganded RAR/RXR led to the appearance of two major DNase I-hypersensitive sites mapping within the RARβ2 promoter (Fig. (Fig.5A,5A, lanes 7 to 9; the new sites are indicated by large horizontal arrows). On addition of RA, the DNase I hypersensitivity generated by the wild-type receptors became more pronounced, mainly for the intensification of the lower signal. Moreover, an additional DNase I-hypersensitive region with lower intensity appeared further upstream of the promoter (small horizontal arrow), while one signal present in the control oocytes disappeared (marked by an asterisk in Fig. Fig.5A;5A; compare lanes 7 to 9 with lanes 10 to 12). It therefore appears that binding of RAR/RXR to chromatin-assembled DNA leads to some nucleosomal structure disruption that, in accordance with the supercoiling assays, becomes more evident following RA treatment. In the presence of unliganded PML-RAR (Fig. (Fig.5A,5A, lanes 13 to 15), only one weaker DNase I-hypersensitive site appeared within the promoter region (Fig. (Fig.5A,5A, lanes 13 to 15). Upon hormone treatment (Fig. (Fig.5A,5A, lanes 16 to 18), the DNase I digestion profile remained almost identical, and no further modifications were observed in flanking regions, suggesting that RA did not provoke any further chromatin disruption. Since binding of PML-RAR to its recognition sequence is significantly improved by RXR (24), we investigated whether coexpression of PML-RAR and RXR changes the DNase I cleavage pattern. Addition of RXR to PML-RAR induced the appearance of the two hypersensitive sites specified by the wild-type receptors (Fig. (Fig.5B,5B, lanes 3 to 6, large horizontal arrows). However, in the presence of PML-RAR/RXR, rather than RAR/RXR, the chromatinized templates remained less accessible to DNase I (Fig. (Fig.5B,5B, lanes 3 and 4 and lanes 5 and 6). Altogether, these results indicate that liganded PML-RAR is less efficient than RAR/RXR in opening the chromatin structure.
We performed similar experiments with MNase. In vivo-assembled chromatin was digested with MNase and hybridized with a promoter-specific (Fig. 6A, C, and E) or a vector control (Fig. 6B, D, and F) probe. Under basal conditions, the RARβ2 promoter was assembled into a well-defined nucleosomal array, which was undistinguishable from the chromatin pattern of vector sequences (Fig. 6A and B, lanes 1 to 3) and not influenced by RA treatment (Fig. 6A and B, lanes 4 to 6). In the presence of unliganded RAR/RXR (Fig. (Fig.6A,6A, lanes 7 to 9), the basal MNase ladder was replaced by a specific (see control hybridization of Fig. Fig.6B)6B) pattern characterized by a smear, particularly evident between the mononucleosome and dinucleosome, and the appearance of a subnucleosomal particle (asterisk). RA treatment led to a specific increase in the subnucleosomal material, suggesting a more substantial alteration of the nucleosomal organization (Fig. (Fig.6A,6A, compare lanes 7 to 9 with lanes 10 to 12). Hybridization with the vector DNA proved that this remodeling activity is confined to RARβ2 regulatory sequences (Fig. (Fig.6B).6B). We next examined the capacity of PML-RAR to remodel the promoter architecture. In the presence of the unliganded fusion protein, the regular nucleosomal array was maintained over the promoter (Fig. (Fig.6C,6C, lanes 7 to 9) and was not modified by RA (Fig. (Fig.6C,6C, lanes 10 to 12). Concomitant expression of RXR and PML-RAR led to the accumulation of a subnucleosomal fragment specific for the RARβ2 sequences (compare lanes 7 to 9 in Fig. 6E and F). However, we did not detect any significant accumulation of the hybridization signal (smear) between the mononucleosome and the dinucleosome. Furthermore, the digestion profile remained unchanged in the presence of liganded PML-RAR/RXR (Fig. 6E and F, lanes 10 to 12). Since the subnucleosomal particle is thought to result from a defined DNA-protein complex (1) and coexpression of RXR increases DNA binding of PML-RAR, we propose that the appearance of this fragment is due to association of nuclear receptors with DNA, as already suggested by Dilworth and colleagues (7). Therefore, unliganded RAR/RXR is able to induce a more significant chromatin disruption than PML-RAR; moreover, only the wild-type nuclear receptors respond to RA, accumulating more subnucleosomal particles, probably because of the recruitment on the promoter of a significant quantity of coactivators.
This hypothesis is reinforced by the MNase profile obtained by digesting chromatin assembled in the presence of both RA and TSA (Fig. (Fig.7).7). The two drugs did not perturb the nucleosomal array assembled over the RARβ2 and the vector sequences (Fig. 7A and B, compare lanes 4 to 6 with lanes 1 to 3). In contrast, in the presence of RAR/RXR (lanes 7 to 9) or PML-RAR/RXR (lanes 13 to 15), the chromatin structure that resulted was disorganized, as shown mainly by the accumulation of the subnucleosomal particle. However, expression of wild-type RAR/RXR (Fig. (Fig.7A,7A, lanes 7 to 9) led to a more dramatic remodeling effect, characterized by a more significant reduction of the mononucleosome and dinucleosome fragments (compare lanes 7 to 9 with lanes 13 to 15). Even in the presence of the two drugs, PML-RAR per se did not show any significant remodeling activity (lanes 10 to 12).
To conclude, the reduced ability of PML-RAR to remodel chromatin might be the consequence of an impaired ability of the fusion protein to recruit coactivators and/or the PolII transcriptional apparatus.
Ectopic expression of the PML-RAR fusion protein into hemopoietic precursors blocks terminal differentiation in vitro and causes leukemias in mice (4, 11, 12). The differentiation block is thought to be the consequence of the aberrant repressive activity of the chimeric protein on RA target genes. In fact, whereas physiological concentrations of RA induce NcoR/HDAC dissociation from the wild-type RAR/RXR heterodimer, only pharmacological concentrations of the same ligand permit its dissociation from PML-RAR (13, 14, 19). Moreover, the ability of PML-RAR to interact with the HDAC complex is critical for the block of hematopoietic differentiation (13). Treatment of AML-derived cell lines or fresh blasts from AML patients with HDAC inhibitors, however, only partially relieves the differentiation block (25), suggesting that other molecular mechanisms may contribute to the transforming ability of PML-RAR. Either other aspects of transcriptional repression or one or more steps involved in the hormone-dependent transcriptional activation observed might also be involved in the onset of APL.
The Xenopus oocyte system is uniquely suited to comparison of the transcriptional properties of RAR and PML-RAR and their influence on chromatin architecture because of the lack of endogenous nuclear receptors and the opportunity to evaluate the role of chromatin in transcriptional regulation. With this system, we have compared the transcriptional effects of RARα and PML-RAR on the immediate-early RA-responsive gene for RARβ2. Moreover, we have analyzed the capabilities of these two proteins to perturb chromatin structure. Our major conclusions are that (i) whereas unliganded RAR/RXR requires chromatin assembly to exert its repressive effects, PML-RAR represses transcription through a further, chromatin-independent, mechanism (Fig. (Fig.2);2); (ii) the activating functions of the fusion protein are impaired compared to those of the wild-type receptor (Fig. (Fig.11 and and2);2); (iii) PML-RAR imposes a more condensed chromatin conformation that remains less accessible even in the presence of RA (Fig. (Fig.33 to to7);7); and (iv) in contrast to those of RAR, the transcriptional and structural functions of PML-RAR do not require heterodimerization with RXR (Fig. (Fig.11 and and33).
Numerous studies have reported the aberrant transcription-repressive activity of PML-RAR and the enhanced association of the fusion protein with the corepressor SMRT/N-CoR complex and suggested that deregulation of chromatin acetylation and altered chromatin organization are important molecular events leading to APL. The results shown in Fig. Fig.11 demonstrate that, indeed, PML-RAR is a much stronger repressor than the wild-type receptor. Since nuclear receptors are virtually absent in oocytes (22, 28), we suggest that the oncoprotein is able to inhibit transcription in the absence of RXR. This finding confirms previous reports showing that the two subunits of RAR/RXR are required to bind one corepressor molecule, whereas PML-RAR per se is able to recruit two SMRT molecules (20). Our analysis of the role played by chromatin in this repression reveals that all of the silencing effect of RAR/RXR depends on nucleosome deposition and histone deacetylation, suggesting that the heterodimer does not target the basal transcription machinery for repression. In the case of PML-RAR, instead, histone deacetylation contributes only partially to repression, and an additional repressive pathway(s) is involved. In fact, strong inhibition is already observed in the absence of a physiological density of nucleosomes (Fig. (Fig.2).2). Since this chromatin-independent repressive effect is not relieved by TSA, deacetylation of a component(s) of the transcriptional apparatus seems not to be involved. The ability of unliganded PML-RAR to directly bind a component(s) of the basal transcriptional machinery, thereby interfering with preinitiation complex formation and/or the interaction of the fusion protein with other corepressor complexes, might explain the additional repressive function of PML-RAR. It is interesting that Xenopus TR/RXR also has a chromatin-independent repressive pathway (29), and in accordance with this, the human TR has been reported to repress transcription by inhibiting the formation of a functional preinitiation complex (10). However, a deacetylase activity seems also to be involved in the transcriptional repression exerted by unliganded TR/RXR when chromatin assembly is far from complete (29).
Even though we cannot rule out the possibility that oligomeric PML-RAR exerts greater steric hindrance on the promoter, we believe that all of our structural data (Fig. (Fig.33 to to6),6), as well as in vivo footprints (S. Segalla et al., unpublished observations), indicate that this is not the case (17).
It has been demonstrated that PML-RAR has a dual function in APL blasts: it contributes to the transformed phenotype and serves as a mediator of retinoid-induced differentiation (11). In fact, while in the presence of near-physiological concentrations of RA, PML-RAR represses transcription and blocks differentiation, large doses of RA activate PML-RAR-mediated gene transcription and cell differentiation (8). In the oocyte system, PML-RAR responds to hormone treatment, allowing, however, a final level of RARβ2 transcriptional activation that is significantly reduced compared to that of the wild-type receptor.
Previous publications have demonstrated that the events underlying transcriptional activation by liganded nuclear receptors are both temporally ordered and interdependent. ATP-dependent remodeling activity and HAT activities are involved, and these chromatin-remodeling steps are thought to create the proper environment for preinitiation complex formation (7, 15, 30). Subsequently, liganded receptors may bind complexes crucial for recruitment of the PolII holoenzyme.
Since incubation of PML-RAR-expressing oocytes with both RA and TSA does not bring RARβ2 expression to levels comparable to those obtained with RAR/RXR (Fig. (Fig.2A),2A), and big differences in activation are observed even when nucleosome deposition is still incomplete (Fig. (Fig.2C),2C), we propose that the fusion protein is impaired in the recruitment of one or more coactivators and/or the transcriptional apparatus. This model is supported by the DNase I indirect end-labeling experiments (Fig. (Fig.5)5) and MNase digestion (Fig. (Fig.66 and and7).7). In fact, the DNase I experiment demonstrated that, in the presence of RA, the wild-type nuclear receptor, but not PML-RAR, specifies new hypersensitive sites, which is indicative of the formation of different nucleoprotein complexes. Likewise, in the MNase assay, only RAR/RXR responded to the hormone, leading to the accumulation of a subnucleosomal DNA fragment, a hallmark of the formation of a DNA-protein complex different from the nucleosome. The reduced ability of PML-RAR to induce the formation of “activating” nucleoprotein complexes is also evident when MNase digestion was performed on oocytes incubated with TSA and RA (Fig. (Fig.7).7). The opportunity to overexpress exogenous proteins in oocytes should help, in the near future, to elucidate which is the limiting step. It is worth mentioning that TBP overexpression failed to restore a complete transcriptional activation response (unpublished data).
It has been proposed that the ability of PML-RAR to recruit nonphysiological concentrations of corepressors on target promoters leads to a chromatin configuration refractory to activating signals (20, 24). We have investigated if the chromatin structure organized on an RARβ2 template differs in the presence of the oncogenic or wild-type receptor. Our results (Fig. (Fig.33 to to5)5) demonstrate that association of unliganded RAR/RXR with nucleosomal DNA produces DNase I-hypersensitive sites indicative of local disruption of histone-DNA contacts. However, the overall chromatin structure, analyzed with a supercoiling assay, is not modified by nuclear receptors. On the contrary, in the presence of PML-RAR, the increase in negative superhelicity demonstrates that minichromosomes have a more compact conformation, which may impede transcription (Fig. (Fig.33 and and4).4). The analysis of DNase I-hypersensitive sites confirmed that chromatin assembled in the presence of unliganded PML-RAR is less accessible (Fig. (Fig.5).5). These results are consistent with the transcriptional data, indicating a more powerful inhibitory activity of the oncogenic protein.
Addition of RA to RAR/RXR-containing oocytes induces chromatin remodeling, evident as a reduction in topological constraints and an increase in DNase I susceptibility in both the promoter and flanking regions. This hormone-dependent chromatin disruption depends on the presence of RXR. In a supercoiling assay, liganded PML-RAR also leads to chromatin disruption; this remodeling activity is facilitated by, but not dependent on, the presence of the coreceptor. However, the final level of negative supercoiling, as well the analysis of DNase I sensitivity, suggests that even in the presence of RA, chromatin assembled in the presence of PML-RAR is more compact and less accessible to transacting factors. These results are in agreement with the expression data.
TSA addition to RAR/RXR-expressing oocytes has the same topological effects as RA. The HDAC inhibitor causes major topological changes in minichromosomes assembled in the presence of PML-RAR; in fact, the overall chromatin structure analyzed with a supercoiling assay is indistinguishable from the one organized in the presence of the wild-type heterodimer, suggesting that the more condensed chromatin configuration organized in the presence of PML-RAR might depend on its ability to recruit aberrant concentrations of HDACs.
Recently, it has been demonstrated that SWI/SNF mediates transcriptional activation and chromatin remodeling with altered DNA topology by activated androgen receptor and thyroid hormone receptor. Artificial SWI/SNF recruitment to chromatin causes a remodeling event comparable to that seen during nuclear-receptor-driven activation but does not lead to any transcriptional induction (15). These results suggest that chromatin remodeling might be a prerequisite necessary, but not sufficient, for transcriptional activation. Accordingly, our experiments indicate that topological variations do not always coincide with the expression data. In fact, TSA and RA induce the same topological effects (Fig. (Fig.3)3) but different quantitative transcriptional results (Fig. (Fig.2).2). Since previous publications (15, 29) and our data (Fig. (Fig.3D)3D) demonstrate that hyperacetylated histones have no effects on chromatin topology and histone acetylation is important for the recruitment of cofactors required for hormone-dependent activation (15), we believe that the topological changes observed in the presence of the exogenous factors and TSA must be the hallmark of the recruitment of new nucleoprotein complexes on the microinjected DNA. The formation of these complexes is strictly dependent on the presence of nuclear receptors but is facilitated by histone acetylation; this might be due to the ability of acetylation to destabilize both local and higher-order chromatin folding and/or to the ability of acetylated histone tails to serve, in accordance with the histone code hypothesis, as docking sites for proteins.
The MNase digestion data shown in Fig. Fig.66 and and77 suggest that binding of unliganded heterodimers, either wild type or oncogenic, leads to a localized chromatin remodeling encompassing the RARβ2 regulatory sequences. However, whereas the wild-type nuclear receptors respond to the ligand augmenting chromatin disruption, liganded PML-RAR/RXR does not lead to any further remodeling. Eventually, PML-RAR alone is never able to modify the nucleosomal distribution.
We thank Luciano Di Croce and Mario Faretta for helpful discussion.
This work was supported by grants from the AIRC and the Cofinanziamento MURST-Università dell'Insubria. Simona Segalla was supported by an AIRC fellowship.