Aberrant Association of Promyelocytic Leukemia Protein-Retinoic Acid Receptor-α with Coactivators Contributes to Its Ability to Regulate Gene Expression*

The aberrant association of promyelocytic leukemia protein-retinoic acid receptor-α (PML-RARα) with corepressor complexes is generally thought to contribute to the ability of PML-RARα to regulate transcription. We report here that PML-RARα acquires aberrant association with coactivators. We show that endogenous PML-RARα interacts with the histone acetyltransferases CBP, p300, and SRC-1 in a hormoneindependent manner, an association not seen for RARα. This hormone-independent coactivator binding activity requires an intact ligand-binding domain and the NR box of the coactivators. Confocal microscopy studies demonstrate that exogenous PML-RARα sequesters and colocalizes with coactivators. These observations correlate with the ability of PML-RARα to attenuate the transcription activation of the Notch signaling downstream effector, CBF1, and of the glucocorticoid receptor. This includes attenuation of the glucocorticoid-induced leucine zipper (GILZ) and FLJ25390 target genes of the endogenous glucocorticoid receptor. Furthermore, treatment of NB4 cells with all-trans-retinoic acid, which promotes PML-RARα degradation, resulted in increased activation of GILZ. On the basis of these findings, we propose a model in which the hormone-independent association between PML-RARα and coactivators contributes to its ability to regulate gene expression.

Acute promyelocytic leukemia (APL) 3 is a disease in which a terminal differentiation block of myeloid precursors occurs at the promyelocytic stage of development (1). APL pathogenesis has been attributed to aberrant signaling due to a common chromosomal translocation involving the retinoic acid receptor-␣ (RAR␣) gene on chromosome 17q21 (2). This chromosomal translocation results in two reciprocal fusion genes that are translated into reciprocal fusion proteins found to be oncogenic. There are five known proteins that create fusions with RAR␣: promyelocytic leukemia (PML), promyelocytic leukemia zinc finger (PLZF), nuclear matrix-associated (NuMA), nucleophosmin (NPM), and signal transducer and activator of transcription 5␤ (Stat5␤) (2)(3)(4)(5)(6)(7). Among them, Ͼ90% of APL patients express a fusion with PML to generate RAR␣-PML and PML-RAR␣.
Disruption of the retinoid signaling pathway is a key pathogenic feature of APL (8 -10). RARs are members of the nuclear receptor family that controls processes such as development, differentiation, and homeostasis through regulation of complex gene networks. RARs form heterodimers with retinoid X receptors (RXRs) and bind to DNA sequences harboring direct repeats, (A/G)G(G/T)TCA, separated by 5 bp. Transcription regulation by RXR/RAR heterodimers involves the exchange of corepressor and coactivator complexes, which are controlled by the hormone binding status of the receptor (11,12). Unliganded RXR/RAR heterodimers bind corepressor complexes to inhibit transcription, whereas hormone-bound RXR/RAR heterodimers dissociate from the corepressors and concomitantly recruit the coactivator complexes, leading to transcription activation. The best known corepressors include SMRT and the nuclear receptor corepressor, which form complexes with mSin3A and histone deacetylases (13,14), leading to deacetylation of histone tails to generate compact chromatin, thus inhibiting the general transcription machinery from reaching the promoter. The coactivators include members of the p160 family, SRC-1, ACTR/RACK1/pCIP/SRC-3, and TIF2/GRIP1, as well as the potent histone acetyltransferases PCAF and p300/ CBP (15)(16)(17)(18)(19)(20)(21)(22). In contrast to the action of histone deacetylases, histone acetyltransferases acetylate histone tails, resulting in weaker associations of histones with chromatin and creating a local environment conducive for binding of the general transcription machinery to the target promoter. These notions are consistent with the observations that histone acetylation is linked to transcription activation and that deacetylation is associated with transcription repression.
Molecular and structural studies of the nuclear receptor ligand-binding domain (LBD) of RXR␣ indicate that, in the absence of hormone, helix 12 of the LBD is held in a position extended away from the rest of the LBD (23). In this conformation, corepressors bind receptors via a common motif, (I/L)XX(I/V)I (where X is any amino acid), which is known as the CoRNR box (24 -26). Hormone binding induces a conformational change in which helix 12 folds back to contact helixes 3-5 of the LBD (23,27). Consequently, such conformational changes result in dissociation of the corepressors and recruitment of the coactivators via the LXXLL motif present in the coactivators (28,29). It has been proposed that these consensus motifs signify a common mechanism of transcription regulation throughout the nuclear receptor family members.
The mechanisms underlying transcription regulation by PML-RAR␣ appear to be more complex than those of RXR␣/ RAR␣ heterodimers. In vitro studies have shown that, in addition to its ability to form heterodimers with RXR␣, PML-RAR␣ is capable of forming homodimers (30,31) that bind corepressors more tightly than does RXR␣/RAR␣ and that require retinoic acid concentrations higher than physiological levels to dissociate from corepressors (Ն1 M) (2,5,31). Therefore, at physiological concentrations of retinoic acid (10 -100 nM), PML-RAR␣ acts as a dominant-negative inhibitor of RARs, leading to a constitutive repression of RAR target genes. This model is supported by transient transfection assays, in which PML-RAR␣ can potently repress transcription when it is tethered to the yeast Gal4 DNA-binding domain (32). There is also evidence suggesting that PML-RAR␣ can promote formation of a heterochromatin-like structure to keep target genes repressed through recruitment of histone methyltransferases such as SUV39H1 (33). However, there is no in vivo evidence to suggest that PML-RAR␣ homodimers compete with RXR␣/RAR␣ for DNA binding. Furthermore, recent reports suggested that overexpression of PML-RAR␣ leads to a reduction of nuclear receptor corepressor protein levels (34,35). Likewise, other studies have indicated that, in the absence of hormone, PML-RAR␣ activates retinoic acid response element-mediated reporter activity in the absence of retinoic acid (36,37). This differential regulation of transcription is likely promoter-and cell type-specific; however, it highlights the fact that a simple model of constitutive repression of RAR␣ target genes by PML-RAR␣ is not sufficient to explain all of its possible roles in transcription regulation.
Although the aberrant association of PML-RAR␣ with corepressors is well characterized, interactions with coactivators have not been well explored, although one study with NPM-RAR␣ suggests that this fusion protein may interact with coactivators in an aberrant manner compared with wild-type RXR/ RAR␣ (38). In this study, we show that PML-RAR␣ homodimers acquire the ability to interact with coactivators in a hormone-independent manner. Consistent with these data, we found that ectopic expression of PML-RAR␣ in the absence of hormone is able to induce mislocalization of some coactivators to a non-nuclear compartment. We also found that over-expression of PML-RAR␣ decreases transcription activation by the glucocorticoid receptor (GR) and CBF1. In light of these findings, we propose a distinct mechanism by which PML-RAR␣ regulates transcription not only as a repressor of RAR␣ target genes, but as a factor capable of repressing the activation of a subset of genes whose activation relies on coactivator recruitment.
Yeast Methods-Yeast two-hybrid assays were carried out according to the protocol of Clontech. Where indicated, yeast cells were grown overnight (16 -24 h) in medium containing all-trans-retinoic acid (ATRA). Liquid ␤-galactosidase activity was assayed, and the values were derived from duplicate experiments with two independent clones.
Luciferase Reporter Assays-Transient transfection assays were carried out according to our published protocol (43). CV-1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 50 units/ml penicillin G, and 50 g/ml streptomycin sulfate at 37°C in 7% CO 2 . For reporter assays, cells (85-95% confluence, 48-well plates) were cotransfected with equal amounts of the corresponding pCMX constructs, not exceeding 1 g, as well as 100 ng of pCBF1-TK-Luc, pGRE-TK-Luc, pERE-TK-Luc, or pMEF2-TK-Luc and 100 ng of pCMX-LacZ in 200 l of Opti-MEM I using the transfection reagent Lipofectamine 2000 following a procedure adapted from Invitrogen. CV-1 cells were transfected, washed, and placed in Dulbecco's modified Eagle's medium containing 10% charcoal-stripped fetal bovine serum, 50 units/ml penicillin G, and 50 g/ml streptomycin sulfate. After 24 h, the medium was replaced or not with 1 M dexamethasone for GR assays or with 1 M estradiol for estrogen receptor-␣ assays or grown in regular Dulbecco's modified Eagle's medium and non-stripped fetal bovine serum for TAN-1 assays. Cells were harvested and assayed for luciferase and ␤-galactosidase activities 36 -48 h after transfection. Luciferase activity was normalized to the level of ␤-galactosidase activity. Each transfection was performed in triplicate and repeated at least three times.
Electroporation of HL-60 cells-HL-60 cells were grown in RPMI 1640 medium supplemented with 10% charcoal-stripped bovine calf serum, 50 units/ml penicillin G, and 50 g/ml streptomycin sulfate at 37°C in 7% CO 2 . For transfection, 7-10 ϫ 10 6 cells were spun down and resuspended in 200 l of Opti-MEM I. 60 g of hemagglutinin (HA)-PML-RAR␣ expression plasmid was added, and the mixture was placed into a 0.2-mm gap electroporation cuvette. The cells were electroporated with a 140-V square-wave pulse for 0.25 ms using the Bio-Rad Gene Pulser system. Following this, the cells were grown 24 h in RPMI 1640 medium and then harvested for immunofluorescence microscopy.
Immunofluorescence Microscopy-CV-1 and HL-60 cells were transfected with an expression plasmid of HA-PML-RAR␣. Transfected cells were grown in medium containing stripped serum for an additional 24 -48 h. For immunofluorescence microscopy, transfected cells were fixed in 3.7% paraformaldehyde in 1ϫ PBS for 30 min at room temperature and permeabilized in 1ϫ PBS with the addition of 0.1% Triton X-100 and 10% goat serum for 10 min. The cells were washed three times with 1ϫ PBS and incubated in a solution of PBS, 10% goat serum, and 0.1% Tween 20 (buffer A) for 60 min. Incubation with primary antibodies was carried out for 120 min in buffer A. The cells were washed three times with 1ϫ PBS, and the secondary antibodies were added for 30 -60 min in the dark at room temperature in buffer A. Coverslips were mounted on slides using VECTASHIELD mounting medium with 4Ј,6-diamidino-2-phenylindole (H-1200, Vector Laboratories). Imaging was performed on a Leica Model DMLB microscope, and pictures were taken with a SPOT camera using SPOT Advanced software (Diagnostic Instruments, Inc.). The pri-mary antibodies used were as follows: mouse anti-HA monoclonal, rabbit anti-p300 polyclonal (N-15), rabbit anti-CBP polyclonal (A-22), and rabbit anti-SRC-1 polyclonal (M-341) (all from Santa Cruz Biotechnology, Inc.). The secondary antibodies used were Alexa Fluor 594-conjugated anti-mouse and Alexa Fluor 488-conjugated anti-rabbit antibodies (Molecular Probes).
Confocal Microscopy-All confocal images were acquired using a Zeiss LSM 510 inverted laser-scanning confocal microscope. A ϫ63 numerical aperture of a 1.4 oil immersion Plan Apochromat objective was used for all experiments. To investigate the localization of transiently transfected HA-PML-RAR␣, images of Alexa Fluor 488 were collected using a 488-nm excitation light from an argon laser, a 488-nm dichroic mirror, and 500 -550-nm band pass barrier filter. For endogenous coregulator anti-SRC-1, anti-CBP, and anti-p300 antibodies, images of Alexa Fluor 594 were collected using a 633-nm excitation light from a He/Ne2 laser, a 633-nm dichroic mirror, and 650-nm long pass filter. All 4Ј,6-diamidino-2-phenylindole-stained nuclear images were collected using a Coherent Mira-F-V5-XW-220 (Verdi 5W) Ti-sapphire laser tuned at 750 nm, a 700-nm dichroic mirror, and a 390 -465-nm band pass barrier filter.
Glutathione S-Transferase (GST) Pulldown Assays-For GST pulldown assays, in vitro translated HA-RAR␣, HA-PML, HA-PML-RAR␣, HA-PLZF-RAR␣, or HA-PLZF was incubated for 30 min on a nutator at room temperature with GST-tagged ACTR receptor interaction domain (RID)-conjugated glutathione-Sepharose beads in NETN buffer with a mixture of protease inhibitors (Sigma) in the presence or absence of ATRA (10 -1000 nM). After incubation, the beads were washed three times with NETN buffer and collected by centrifugation. The proteins were eluted and denatured by placing the samples at 100°C for 5 min and then run on 10% SDS-polyacrylamide gels. The products were visualized by Western blot analysis with HA-conjugated anti-horseradish peroxidase antibody.
Reverse Transcription (RT)-PCR-A549 cells were transfected with HA, HA-PML-RAR␣, or HA-PML-RAR␣(F584A) using Lipofectamine 2000. Where indicated, samples were treated for 16 -20 h with 1 M dexamethasone before harvesting. One-third of the cells were used to make whole cell lysates using radioimmune precipitation assay buffer plus protease inhibitors. The other two-thirds of the cells were used for RNA isolation. NB4 cells were grown in RPMI 1640 medium supplemented with 10% charcoal-stripped fetal bovine serum, 50 units/ml penicillin G, and 50 g/ml streptomycin sulfate at 37°C in 7% CO 2 . They were either left untreated or treated alone or in combination with 1 M ATRA for 72 h and 1 M dexamethasone for 16 -18 h. RNA isolation was performed using an RNeasy mini RNA isolation kit (Qiagen Inc.). All procedures were performed according to the manufacturer's protocol. DNase I digestion was performed on an RNeasy column following the manufacturer's protocol using RNase-and DNase-free DNase I (Qiagen Inc.). The isolated RNA was used in a semiquantitative RT-PCR using a one-step RT-PCR kit (Invitrogen) according to the manufacturer's protocol. 500 ng of RNA template was used in each reaction, and PCR amplification was repeated for 40 cycles. The final primer concentra-tion in each reaction was 0.2 M. Each assay was performed at least three times. The actin primers used as a control were 5Ј-ggtctcaaacatgatctgggtc-3Ј (forward) and 5Ј-aaatctggcaccacaccttc-3Ј (reverse). The GR target gene primers used were as follows: glucocorticoid-induced leucine zipper (GILZ), 5Ј-agatcgaacaggccatggat-3Ј (forward) and 5Ј-ttacaccgcagaaccaccag-3Ј (reverse); and FLJ25390, 5Ј-ggctcatgctggatgacaa-3Ј (forward) and 5Ј-cccagatggtggagatcagt-3Ј (reverse). 4

RESULTS
The observations that PML-RAR␣ requires higher concentrations of hormone to dissociate from corepressors and that both corepressor-and coactivator-binding domains overlap in the modeled nuclear receptor LBD suggested that PML-RAR␣ may interact with both sets of coregulators differently from wild-type RAR␣ (23). To investigate this possibility, we characterized the association of PML-RAR␣ with coactivators in solution by GST pulldown experiments. A fusion protein composed of GST attached to the RID of SMRT or ACTR was expressed and purified from bacteria. Purified immobilized GST-SMRT-(RID) or GST-ACTR(RID) fusion protein was incubated with in vitro transcribed and translated RAR␣, PML, or PML-RAR␣ in the presence and absence of hormone. To our surprise, PML-RAR␣ could be pulled down by GST-ACTR(RID) in both the presence and absence of hormone (Fig. 1A, lanes 4 and 5). No binding was observed with any of the proteins to GST alone. In contrast and as expected, RAR␣ was able to bind to ACTR in a DNA-independent manner, but only after the addition of hormone (Fig. 1B, lane 4 versus lane 5). Also, PML was unable to bind to GST-ACTR(RID) regardless of the presence of hormone (Fig. 1C, lanes 4 and 5). As a control, Fig. 1D shows that the association of PML-RAR␣ with SMRT in this assay is consistent with the well documented association, which requires 1 M ATRA to induce complete dissociation (lanes [3][4][5]. These results suggest that PML-RAR␣ binds coactivators in solution in both the absence and presence of hormone. To investigate the molecular basis of the hormone-independent association between PML-RAR␣ and coactivators, we performed yeast-two hybrid assays. The yeast Gal4 activation domain (pGAD) was fused to the LBD of RAR␣, PML, or PML-RAR␣, whereas ACTR(RID) was fused to the Gal4 DNA-binding domain (pGBT9). The constructs were cotransformed into yeast. The interaction of ACTR(RID) and PML-RAR␣ was assayed by measuring ␤-galactosidase expression. As shown in Fig. 2A, as controls, coexpression of the pGBT9 vector alone and pGAD-PML-RAR␣ or pGBT9-ACTR(RID) and the pGAD vector alone resulted in only basal reporter activity. Moreover, coexpression of pGBT9-ACTR(RID) and pGAD-PML did not activate reporter activity in the presence or absence of hormone. As expected, coexpression of pGBT9-ACTR(RID) and pGAD-RAR␣(LBD) resulted in activation of the reporter only in the presence of hormone. However, consistent with our in vitro protein-protein interaction data (Fig. 1), coexpression of pGBT9-ACTR(RID) and pGAD-PML-RAR␣ led to a potent activation of reporter activity in the absence of hormone, indicating a strong association between ACTR(RID) and PML-RAR␣. The addition of hormone modestly potentiated this interaction. Taken together, these results indicate that PML-RAR␣ interacts with ACTR(RID) in yeast independently of hormone.
To test whether this interaction is unique only to ACTR, we performed a similar yeast two-hybrid analysis using constructs bearing the RID from SRC-1 and GRIP1. Fig. 2B demonstrates that, similar to ACTR(RID), RAR␣(LBD) associated only with SRC-1(RID) and GRIP1(RID) in the presence of hormone, whereas PML-RAR␣ was able to associate with coactivators in both the presence and absence of hormone. To investigate whether this association extends to other histone acetyltransferases as well as the p160 family, yeast two-hybrid assays were performed to map the PML-RAR␣ interaction domain within CBP (Fig. 2C). Our results show that, although coexpression of the pGAD vector and pGBT9-CBP-(2-460) resulted in a slight activation of reporter activity (first bar), pGBT9-CBP-(2-460) and pGAD-PML-RAR␣ further activated expression of reporter activity (second bar), indicating an interaction between PML-RAR␣ and CBP-(2-460). The region of CBP involved in the interaction (residues 2-460) includes two LXXLL motifs at amino acids 68 -78 and 355-365, which have previously been shown to be involved in associating with RAR␣ (44). These data demonstrate that the hormone-independent interaction between PML-RAR␣ and coactivators occurs with all three p160 coactivators tested as well as with CBP.
Because PML-RAR␣ associated with the coactivators in a hormone-independent manner, a characteristic not shared with RAR␣, it is possible that the actual binding between PML-RAR␣ and coactivators is distinct from that of RAR␣. To investigate this possibility, we performed yeast two-hybrid assays A, PML-RAR␣ associates with ACTR in both the presence and absence of hormone. PML-RAR␣ binding to GST-ACTR(RID) was measured by GST pulldown assays. Immobilized purified GST-ACTR(RID) was incubated with in vitro translated HA-PML-RAR␣ in the absence or presence of 1 M ATRA. Bound complexes were eluted and analyzed by Western blot analysis with anti-HA antibody. B, RAR␣ only associates with ACTR(RID) in the presence of hormone. HA-RAR␣ binding to GST-ACTR(RID) was monitored by GST pulldown assays as described for A using in vitro translated HA-RAR␣. C, PML does not bind ACTR(RID). HA-PML binding to GST-ACTR(RID) was monitored by GST pulldown assays as described for A using in vitro translated HA-PML. D, PML-RAR␣ binding to SMRT requires 1 M ATRA to induce dissociation. HA-PML-RAR␣ binding to GST-SMRT(RID) was monitored by GST pulldown assays as described for A using increasing amounts of ATRA (10 nM to 1 M).
using mutant ACTR constructs. There are three LXXLL-containing coactivator-binding motifs in ACTR(RID) (45). Each construct contained a single, double, or triple mutation of these three motifs, numbered 1-3 from the N terminus to C terminus. In the absence of ATRA, no interaction was observed (data not shown). In the presence of ATRA, we found that mutation of each of the motifs decreased the interaction with RAR␣. However, mutation of the second motif had the most severe effects on the association (Fig. 3A). The interaction was abolished in any of the constructs with a mutation of the second motif. A similar assay was performed for PML-RAR␣, except that yeast cells were grown in the absence of hormone. Fig.  3B shows that, similar to RAR␣, although all of the mutations decreased the association compared with the wild-type construct, mutations of the second motif abrogated the hormone-independent interaction. These data indicate that both the ligand-independent association of PML-RAR␣ with ACTR and the ligand-dependent association of RAR␣ with ACTR are mediated through LXXLL-containing motifs in the coactivators.
Although these data indicate that the hormone-independent association of coactivators with PML-RAR␣ occurs in a manner similar to the hormone-dependent association with RAR␣, we hypothesized that there may also be contributions to this interaction due to the PML portion of PML-RAR␣, which is absent in RAR␣. To investigate this possibility, we generated PML-RAR␣ constructs harboring a deletion of one functional region of the RBCC (ring finger/B boxes/coiled-coil domain) motif located in the N terminus of PML. These deletion constructs were tested for their ability to bind to PML-RAR␣ in a GST pulldown assay similar to that described for Fig. 1. In vitro translated PML-RAR␣ proteins were incubated with immobilized GST-ACTR(RID) to compare the ability of the deletion mutants to associate with ACTR in the absence of hormone with that of wild-type PML-RAR␣. Interestingly, the deletion of PML regions did not seem to have any significant effect on the binding of PML-RAR␣ to ACTR (Fig. 3C, lane 1 versus lanes [2][3][4][5]. These data suggest that the LBD of the RAR␣ portion of PML-RAR␣ is the major contributor to the hormone-independent interactions with coactivators. To confirm this, we tested the binding of ACTR to a PML-RAR␣ mutant containing a mutation at a site shown previously to abolish the interaction of RAR␣ with coregulators in vivo by a mammalian two-hybrid assay (46). We found that F584A (corresponding to F249A in RAR␣) completely abolished the interaction of PML-RAR␣ with ACTR in comparison with wild-type binding (lane 6). To further analyze the activity of this mutant, we also confirmed by yeast two-hybrid analysis that the interaction of PML-RAR␣(F584A) was significantly less than that of wild-type PML-RAR␣ with ACTR (Fig. 3D). These data show that, similar to the results from the yeast two-hybrid analysis of LXXLL motifs, the hormone-independent interaction between PML-RAR␣ and coactivators appears to be mediated similarly to the hormone-dependent interaction between RAR␣ and coactivators.
One main difference between PML-RAR␣ and RAR␣ is the ability of PML-RAR␣ to form homodimers, a characteristic common to all fusion proteins expressed in APL. To test Binding of PML-RAR␣ to ACTR with or without 1 M ATRA was assayed by yeast-two hybrid assays. Yeast expression plasmids pGBT9-ACTR(RID) and pGAD-PML-RAR␣, pGAD-RAR␣, or pGAD-PML were cotransformed into yeast strain Y190. Liquid ␤-galactosidase assays were performed according to the Clontech protocol. ␤-Galactosidase (␤-gal) activity was normalized by total cell numbers and time. AD, activation domain. B, PML-RAR␣ associates with other p160 family members in a hormone-independent manner. Binding of PML-RAR␣ and RAR␣ to SRC-1 and GRIP1 in the presence or absence of hormone was measured by yeast two-hybrid assays as described for A using pGBT9-SRC-1(RID) or pGBT9-GRIP1(RID) and either pGAD-RAR␣ or pGAD-PML-RAR␣. C, PML-RAR␣ associates with CBP in the absence of hormone. Binding of PML-RAR␣ to CBP was assayed by yeast two-hybrid assays with pGAD-PML-RAR␣ and different pGBT9-CBP constructs containing fragments of CBP as described for A.
whether this hormone-independent association with coactivators extends to other fusion proteins expressed in APL and therefore may be somewhat dependent on the homodimerization of the fusion protein, we tested the ability of PLZF-RAR␣ to associate with ACTR by GST pulldown assays. Similar to PML-RAR␣, PLZF-RAR␣ was capable of associating with ACTR in both the presence and absence of hormone (Fig. 3E). As a control, no interaction between PLZF and ACTR(RID) was observed under these conditions (Fig. 3F). Furthermore, this ligand-independent association was confirmed using a yeast-two hybrid assay (Fig. 3G). These data further support the idea that the hormone-independent association of PML-RAR␣ with coactivators involves residues in the RAR␣ portion of the fusion protein and is similar to that observed for the hormone-dependent association of RAR␣ with coactivators.
To confirm that this hormone-independent association occurs in mammalian cells as well as in vitro, we performed co-immunoprecipitation studies. HA-PML-RAR␣ was coexpressed with FLAG-ACTR in mammalian cells. Immunoprecipitations against the FLAG epitope and subsequent Western blotting with anti-FLAG and anti-HA antibodies were per-FIGURE 3. Hormone-independent association of PML-RAR␣ with coactivators occurs via the same regions as hormone-dependent association. A, the second LXXLL motif of ACTR is the most important in the hormone-dependent association with RAR␣. A general schematic of ACTR is shown indicating the three LXXLL motifs (M1-M3) in the RID that were mutated. Yeast two-hybrid assays were performed using wild-type (WT) and mutant pGAD-RAR␣ and pGBT9-ACTR constructs as described for Fig. 2A. B, the second LXXLL motif of ACTR is the most important in the hormone-independent association with PML-RAR␣. Yeast two-hybrid assays were performed using wild-type and mutant pGAD-PML-RAR␣ and pGBT9-ACTR constructs as described for Fig. 2A. C, mapping the domains critical for the hormone-independent association with ACTR. Wild-type and mutant HA-PML-RAR␣ binding to GST-ACTR(RID) was monitored by GST pulldown assay as described for Fig. 1A using in vitro translated proteins for HA-PML-RAR␣, HA-PML-RAR␣(⌬RING), HA-PML-RAR␣(⌬BBox1) (⌬B1), HA-PML-RAR␣(⌬BBox2) (⌬B2), HA-PML-RAR␣(⌬coiled coil) (⌬CC)), and HA-PML-RAR␣(F584A). The ratio of the amount of binding of each mutant compared with that of the wild-type protein is indicated. Coomassie Blue staining is shown to indicate equal loading for the GST-beads in each lane. D, PML-RAR␣(F584A) does not interact with ACTR in yeast. Yeast two-hybrid assays were performed in the absence of hormone as described for Fig. 2A using wild-type or mutant pGAD-PML-RAR␣. AD, activation domain; ␤-gal, ␤-galactosidase. E, PLZF-RAR␣ interacts with ACTR in a hormone-independent manner. HA-PLZF-RAR␣ binding to GST-ACTR(RID) was monitored by GST pulldown as described for Fig. 1A using in vitro translated protein. F, PLZF alone does not interact with ACTR(RID). HA-PLZF binding to GST-ACTR(RID) was monitored by GST pulldown as described for Fig. 1A using in vitro translated protein. G, PLZF-RAR␣ interacts with ACTR in yeast. Yeast two-hybrid assays were performed using the pGAD-PLZF-RAR␣, pGAD-RAR␣, pGAD-PLZF, and pGBT9-ACTR constructs as described for Fig. 2A.
formed on whole cell lysates to examine whether ACTR forms a complex with PML-RAR␣ (Fig. 4A). Our results show that HA-PML-RAR␣ was detected in the immunopellet only when FLAG-ACTR was coexpressed, indicating an interaction between these two proteins. To determine whether this extends to other members of the coactivator complex, we performed a similar co-immunoprecipitation study with FLAG-PCAF coexpressed with either HA-PML-RAR␣ or HA-RAR␣ as a control (Fig. 4B). In this assay, immunoprecipitation was performed in both the presence and absence of ATRA. We found that, as expected, HA-RAR␣ co-immunoprecipitated only with FLAG-PCAF in the presence of hormone (lane 12 versus lane 14), whereas HA-PML-RAR␣ was able to co-immunoprecipitate with FLAG-PCAF in both the absence and presence of hormone (lanes 11 and 13). To examine whether endogenous PML-RAR␣ associates with coactivators, we prepared extracts from NB4 cells, which endogenously express PML-RAR␣. We examined whether coactivators are expressed in NB4 cells by immunoprecipitation followed by Western blotting. We found that the coactivators CBP, p300, and SRC-1 are expressed in NB4 cells (Fig. 4C). We then used an antibody against RAR␣ to test whether it can precipitate PML-RAR␣. Fig. 4D demonstrates that this antibody was able to detect both RAR␣ and PML-RAR␣ (lanes 1-3 and data not shown) in these lysates. Subsequent stripping of this membrane and reprobing with anti-PML antibody confirmed that the higher molecular mass band detected with anti-RAR␣ antibody is PML-RAR␣ (lanes 4 -6). To investigate whether the coactivators interact with PML-RAR␣, we performed co-immunoprecipitation using anti-coactivator antibodies for immunoprecipitation and anti-RAR␣ antibody for Western blotting. Fig. 4E shows that all three coactivators were able to immunoprecipitate endogenous PML-RAR␣. Furthermore, we grew

. Hormone-independent association of PML-RAR␣ with coactivators in mammalian cells.
A, PML-RAR␣ interacts with ACTR in mammalian cells. Immunoprecipitation of ACTR was performed on lysates of 293 cells transfected with FLAG-ACTR and HA-PML-RAR␣. Western blot (WB) analysis was performed on whole cell lysates (WCE) and immunoprecipitated (IP) complexes with anti-FLAG and anti-HA antibodies. B, PML-RAR␣ interacts with PCAF in mammalian cells in both the presence and absence of hormone, whereas RAR␣ interacts only in the presence of hormone. Immunoprecipitation of PCAF was performed as described for A in cells expressing either FLAG-PML-RAR␣ or FLAG-RAR␣. Where indicated, cells were grown and immunoprecipitation was performed in the presence of 1 M ATRA. C, CBP, p300, and SRC-1 are expressed in NB4 cells. Immunoprecipitation was performed with lysates from NB4 cells using the indicated anti-coactivator antibodies. Western blot analysis was performed using the antibodies for immunoprecipitation. D, immunoprecipitation of PML-RAR␣ by anti-RAR␣ antibody in NB4 cells. Western blotting was performed first with anti-RAR␣ antibody (lanes 1-3). The membrane was subsequently stripped and reprobed with anti-PML antibody (lanes 4 -6). E, endogenous PML-RAR␣ associates with SRC-1, p300, and CBP. Immunoprecipitations were carried out as described for D, and the immunopellets were probed with anti-RAR␣ antibody. F, PML-RAR␣ associates with CBP in both the absence and presence of hormone, whereas RAR␣ primarily associates with CBP in the presence of hormone. Immunoprecipitations were carried out as described for D, and the immunopellets were probed with anti-RAR␣ antibody. Where indicated, cells were grown and immunoprecipitations were performed in the presence of 1 M ATRA. IB, immunoblot. For C-F, 5% of the input is shown.
NB4 cells in the presence of ATRA for 12 h and compared the ability of PML-RAR␣ and RAR␣ to immunoprecipitate with CBP in the absence and presence of hormone (Fig. 4F). As documented previously (47), PML-RAR␣ protein was decreased after ATRA treatment (lane 1 versus lane 2); however, anti-CBP antibody was able to efficiently immunoprecipitate PML-RAR␣ under both conditions (lanes 3 and 4). Conversely, as expected, immunoprecipitation of RAR␣ by CBP was readily detectable only in the presence of hormone (lanes 3 and 4). Taken together, these data demonstrate that PML-RAR␣ associates with coactivators in mammalian cells in a hormone-independent manner.
To further characterize the interaction of PML-RAR␣ with coactivators, we carried out immunofluorescence microscopy. To examine the colocalization of PML-RAR␣ and coactivators in a cell line similar to cells endogenously expressing PML-RAR␣, we performed confocal microscopy with HL-60 cells (Fig. 5, A-C). In these cells, exogenous PML-RAR␣ appeared to be localized mostly in the nucleus. Furthermore, PML-RAR␣ partially colocalized with the coactivators SRC-1, CBP, and p300. However, because PML-RAR␣ may not always localize to the nucleus in APL cells, we also tested the effects of PML-RAR␣ on coactivators in CV-1 cells, in which PML-RAR␣ is mostly cytoplasmic. We found that overexpression of PML-RAR␣ in CV-1 cells led to a redistribution of coactivators. For SRC-1 and CBP, staining was mostly nuclear in untransfected cells, but more widespread and colocalized with PML-RAR␣ in transfected cells (Fig. 5D, panels a-f). As a control, transfected RAR␣ localized to the nucleus, and its overexpression in the cells was unable to disrupt the nuclear localization of coactivators (panels g-i). Furthermore, confocal microscopy studies confirmed these phenomena and indicated that, similar to SRC-1 and CBP, immunostaining for p300 showed a change from a mostly nuclear pattern in untransfected cells to a more diffuse pattern resembling that of PML-RAR␣ in transfected cells (Fig. 5E), suggesting that PML-RAR␣ colocalizes with p300. These data are consistent with our protein-protein interaction data, suggesting that PML-RAR␣ interacts with coactivators in the absence of hormone in mammalian cells.
Although PML-RAR␣ is capable of binding to DNA at a retinoic acid response element in vitro and possibly when overexpressed in vivo, there is currently no evidence that PML-RAR␣ acts as a direct transcription factor in APL through direct regulation of RAR target genes. However, because PML-RAR␣ is able to interact with coactivators in a hormone-independent manner, we hypothesized that PML-RAR␣ could affect other genes whose activation involves binding to coactivators by possibly sequestering the coactivators from binding to other transcription factors. To test this hypothesis, we performed transient transfection assays. We first examined the effects of PML-RAR␣ on Notch-mediated transactivation of the CBF1 transcription factor. TAN-1 is an intracellular fragment of the Notch transmembrane receptor that activates CBF1 activity through the recruitment of coactivator complexes, including PCAF and p300/CBP (48). To test whether PML-RAR␣ has effects on the activation of CBF1, we carried out transient transfection experiments using a luciferase reporter gene construct containing a CBF1 response element (40). We found that, as expected, TAN-1 potently activated reporter activity (Fig. 6A,  bar 2). Furthermore, TAN-1-mediated activation of CBF1 was attenuated by PML-RAR␣ (Fig. 6B, bars 3-5); however, PML-RAR␣(F584A), which did not interact strongly with coactivators, did not have significant effects on reporter activity (bars 6 -8). We also investigated whether PML-RAR␣ has effects on activation of a luciferase reporter gene under the control of a GR response element. GR is a member of the family of class I nuclear receptors that bind DNA as homodimers to regulate transcription. In part, their activation of transcription is similar to RXR␣/RAR␣ in that they require hormone to bind a coactivator complex that includes the p160 coactivators and p300/ CBP (49,50). Fig. 6B demonstrates that, in the presence of dexamethasone, GR highly activated transcription of the GR response element-containing reporter construct (bars 1 and 2). The addition of PML-RAR␣ led to decreased reporter activity in a dose-dependent manner (bars 3-6). Also, similar to TAN-1, this effect of PML-RAR␣ was largely dependent on the association of PML-RAR␣ with coactivators because PML-RAR␣(F584A) no longer had any significant effect on GR reporter activation (bars 7-10). Taken together, these results indicate that PML-RAR␣ is able to inhibit the transcription activation of GR and CBF1 and that this inhibition depends on physical interaction with coactivators.
To further examine the effects of PML-RAR␣ on GR transcription, we investigated the effects of exogenous PML-RAR␣ on expression of endogenous GR target genes. GILZ and FLJ25390 are two genes that have been identified as GR targets in A549 cells and that are activated in response to dexamethasone. 5 We analyzed these genes for transcription activity in the absence and presence of PML-RAR␣ by RT-PCR. A549 cells were transfected with HA vector alone, HA-PML-RAR␣, or HA-PML-RAR␣(F584A). Cells were treated with dexamethasone, and the RNA was isolated for RT-PCR to measure expression of GILZ and FLJ25390. We found that PML-RAR␣, but not the vector alone or the F584A mutant, resulted in a decrease in activation of GILZ and FLJ25390 (Fig. 6C, lanes 3 and 4 versus  lanes 1, 2, 5, and 6). Quantification analysis indicated that expression of PML-RAR␣ blocked the hormone-induced expression of GILZ and FLJ25390, whereas expression of the vector alone and PML-RAR␣(F584A) did not have this effect. Furthermore, GILZ was expressed and regulated by dexamethasone in NB4 cells, an APL patient-derived cell line that expresses PML-RAR␣. To test whether PML-RAR␣ affects GILZ gene expression in these cells, we treated the cells with a concentration of ATRA shown previously to induce degradation of PML-RAR␣ (47). RT-PCR analysis of GILZ indicated that dexamethasone treatment resulted in an increase in GILZ expression. In addition, when ATRA was also included with dexamethasone, there was a further increase in GILZ expression (Fig. 6D). These results indicate that the presence of PML-RAR␣, even at endogenous levels, results in a decrease in GR target gene expression. These data are consistent with our model in that, by interacting with coactivators in a hormoneindependent manner, PML-RAR␣ is able to sequester these transcription regulators from associating with transcription factors that rely on them for proper control of transcription.

DISCUSSION
To further define the mechanism of transcription regulation by PML-RAR␣, we undertook a study of the association of PML-RAR␣ with coactivators. PML-RAR␣ is known to have an aberrant hormone responsiveness that results in reduced dissociation of corepressors. To our surprise, we found that unliganded PML-RAR␣ interacted with coactivators both in mammalian cells and in vitro. This result is reminiscent of our previous study in which RXR␣/RAR␣(⌬443), a helix 12 deletion mutant, acquired not only hormone-resistant dissociation with the corepressors, but also hormone-independent association with the coactivators (46). In addition, hormone further promoted the association between PML-RAR␣ and coactivators, as observed in GST pulldown assays (Fig. 1) and electrophoretic mobility shift assays (data not shown). In the context of PML-RAR␣ being a constitutive repressor, this hormone-promoted coactivator binding likely plays a role in the response of patients to pharmacological levels of ATRA by mediating the release of corepressors and the binding of coactivators to PML-RAR␣.
Through the use of our mutation and deletion analysis of PML-RAR␣ to investigate this interaction, we have shown that the region of binding for the hormone-independent association is similar to that for hormone-dependent interactions of PML- . The graphs represent the amount of product obtained from dexamethasone-treated samples over those that were untreated. All bands are relative to the actin levels of each sample as a control. D, treatment of ATRA results in increased activation of an endogenous GR target gene in NB4 cells. RT-PCR was performed on RNA isolated from NB4 cells after treatment with ATRA and/or dexamethasone to analyze expression of GILZ. Quantification of the intensity of RT-PCR products is shown and was analyzed as described for C. RAR␣ and RAR␣ with coactivators. Although deletion of the functional domains of PML in PML-RAR␣ had no drastic effect on the ability to bind to ACTR, the loss of the RING domain did appear to decrease the interaction to ϳ60% that of the wildtype protein; however, how the function of the RING domain may affect this interaction is currently unclear. Furthermore, a mutation in the LBD of RAR␣, found previously to abolish the interactions with coregulators (46), had a similar effect on PML-RAR␣, indicating that the LBD is primarily responsible for this interaction. On the basis of these results, we suggest that, although the main site of interaction between PML-RAR␣ and coactivators in the absence of hormone is the LBD of RAR␣, it is likely that the PML portion of the fusion protein may have some effects on this interaction due to dimerization status and interactions with other proteins in the cell. This hormone-independent association with coactivators is not unique to PML-RAR␣ among APL fusion proteins, as we also observed the association to occur with PLZF-RAR␣. Likewise, a similar hormone-independent interaction between NPM-RAR␣ and ACTR has been observed previously (38). Whether or not this interaction plays a role for all of the different APL fusion proteins remains to be tested, however, because recent evidence suggests that the different fusion partners of RAR␣ result in leukemias with different biological features in a mouse model of APL (51).
Although the structural basis for the interactions between coregulators and PML-RAR␣ is currently undefined, our and other results validate a mechanistic link of the interactions between receptors and the coactivators and corepressors. These studies with both corepressors and coactivators further suggest that, although the LBDs of the fusion proteins are similar to that of wild-type RAR␣, it is likely that there are small conformational changes that are responsible for the differential binding to these coregulators. We propose that the association of PML-RAR␣ with coactivators results in sequestration of these proteins, limiting their availability to bind other transcription factors. This would then result in aberrant transcription of these genes, as indicated by our studies with a reporter gene under the control of either GR or CBF1, whose transcription activation requires association with the coactivators, including GRIP1 and/or p300/CBP. This hypothesis is further supported by our data indicating that ectopic expression of PML-RAR␣ in A549 cells results in a decrease in dexamethasone-induced activation of endogenous GR target genes GILZ and FLJ25390 (Fig. 6C). Although the effects of this overexpression may seem modest at only an ϳ20 -30% reduction in the amount of transcription induced by dexamethasone for PML-RAR␣ compared with that for either the vector alone or the F584A mutant, which has a reduced infinity for coactivators, it is important to note that the RT-PCR was performed on the whole population of cells, not just those transfected. Furthermore, our data in NB4 cells analyzing GILZ expression before and after treatment with ATRA to induce degradation of PML-RAR␣ indicate that this effect is not only a result of PML-RAR␣ overexpression and therefore may play a role in APL leukemogenesis. The effects of PML-RAR␣ on target genes are likely dependent on the environment and the local concentration of PML-RAR␣ around the promoter, as we did not observe an effect of PML-RAR␣ on the transcription activation of either estrogen receptor-␣ or MEF2 (supplemental Fig. 1). We hypothesize that the binding affinity and preference of PML-RAR␣ and the target transcription factor for different coactivators play a role in determining the level to which PML-RAR␣ may affect its transcription activation.
Our model, in which PML-RAR␣ has effects on a wide array of other cellular targets, supports a previous report demonstrating that PML-RAR␣ affects both nuclear receptor as well as non-nuclear receptor signaling in myeloid differentiation (8). Coincidentally, recent microarray data on genes whose expression changed upon retinoic acid treatment of APL blasts or U937 cells expressing either wild-type RAR␣ or PML-RAR␣ reveal that a number of genes show similar patterns of expression in the presence of the two receptors. Furthermore, there are subsets of genes unique to each (52,53), indicating that PML-RAR␣ also has effects on expression of non-RAR␣ target genes and further suggesting that PML-RAR␣ does not simply bind and/or affect only genes regulated by RAR␣.
Determining the exact endogenous localization of PML-RAR␣ is difficult at this point due to the lack of an appropriate cell line and the lack of appropriate antibodies that specifically recognize PML-RAR␣ without cross-reacting with PML or RAR␣. Two of the cell lines employed in this study show differential subcellular distribution. In HL-60 cells, we found a mostly nuclear localization of overexpressed PML-RAR␣, though it was unevenly distributed throughout the nucleus. However, because myoblasts, which express PML-RAR␣, have large nuclei and very little cytoplasm, it is difficult to visualize changes in protein subcellular localization. To avoid this problem and to further support our data, we transiently overexpressed HA-PML-RAR␣ in CV-1 cells, a cell line used throughout the nuclear receptor field and in which movement of proteins between the nucleus and cytoplasm is more easily detected, and performed confocal microscopy to examine any effect overexpressed HA-PML-RAR␣ had on the endogenous coactivators. We found that PML-RAR␣ is localized predominantly in the cytoplasm in this cell line. In either case, a significant amount of colocalization of PML-RAR␣ and coactivators was observed, indicating possible interaction in the absence of hormone. Notably, we have demonstrated that endogenous PML-RAR␣ associates with coactivators in NB4 cells. We hypothesize that the hormone-independent association with coactivators likely plays a role in situations in which PML-RAR␣ may not be completely nuclear, as was reported for APL blasts (54,55), as well as when PML-RAR␣ is likely nuclear due to overexpression of PML-RAR␣ in the cell, enabling it to sequester a large portion of the coregulators (8,56).
Nevertheless, the unique properties of PML-RAR␣ based on our findings will give rise to many questions in redefining its mechanism of action. First, because the binding of corepressors and coactivators to PML-RAR␣ is mutually exclusive and because both species exist in at least some physiological settings, there must be tight regulation of the interplay between these factors in the absence or presence of low levels of hormone. It has been suggested that PML-RAR␣ may act in some circumstances to decrease the physiological concentration of corepressors (34,35). This would then alter the ratio of core-pressors to coactivators, and based on our findings, it may lead to alternate transcription regulation by PML-RAR␣. Second, it is likely that PML-RAR␣ binding of corepressors or possible sequestering of coactivators preferentially affects transcription of some genes over others. In accordance with this idea, it has been suggested that, because PML-RAR␣ disrupts the normal PML localization in PML bodies, it may relocalize cofactors by subsequently disrupting their normal localization, thereby preventing them from functioning normally (57). Third, it is possible that PML-RAR␣ interacts with other transcription factors to modulate transcription by recruiting coactivators or corepressors, similar to one mechanism by which GR has been proposed to affect transcription (49,50). Understanding the role of PML-RAR␣ and deciphering its interactions with coregulators under these different circumstances will help to further determine the altered signaling that leads to the differentiation block.
The current treatments for APL, which involve targeting PML-RAR␣ for degradation and inhibiting the activity of histone deacetylases, can unfortunately have various deleterious effects in the cell due to interference with the wild-type PML and RAR␣ proteins (47, 58 -60). Recent suggestions for treatment also include targeting other oncogenic pathways that may be involved in APL such as the Ras pathway (61), but these would be in addition to the traditional treatments. In the same token, long-term cancer management with the available treatments, the most common being pharmacological levels of ATRA, can lead to hormonal resistance, which impedes the ability of the drugs to help the patient overcome the illness. Finally, although most APL patients express fusions with PML, which are ATRA-responsive, there are forms of PML-RAR␣ as well as other fusion proteins such as PLZF-RAR␣ that are ATRA-resistant. Based on our results and those of others studying interactions of PML-RAR␣ with corepressors, it is possible that there are small changes in the LBD of PML-RAR␣ that allow coregulator associations. This may present possible targets of therapies that could more specifically act on the fusion protein to block its action while leaving other cellular functions intact, thus decreasing side effects and hopefully increasing the potency of treatment and widening the available patient base to better treat all APL patients. Ground work on targeting transcription factor-coregulator associations has been started, and it has been shown that a small peptide can be used to induce dissociation of the corepressor SMRT from Bcl-6 (62, 63), suggesting that, with more knowledge of these properties of PML-RAR␣, new and improved possibilities for patient treatment could be on the horizon.