Reduced Retinoic Acid-Sensitivities of Nuclear Receptor Corepressor Binding to PML- and PLZF-RARα Underlie Molecular Pathogenesis and Treatment of Acute Promyelocytic Leukemia

Constructions of PLZF, 5′PLZF (N-terminal region of PLZF, amino acids 1-455), PLZFΔPOZ (PLZF deleted for the BTB/POZ domain, amino acids 1-120), PML-RARα, PLZF-RARα cDNA expression vectors, and GST-PLZF plasmid have been previously described.40,41Mammalian, bacterial, and in vitro expression vectors for the PML,8 full-length and partial N-CoR,30,42HDAC1,35 as well as Sin3A and B³⁶ proteins were described by others. L4-tkluc reporter construct contains four LexA operators with two PLZF target sites (5′-GTACAGTAC-3′), which were derived using PCR from L8-CAT plasmid43 and cloned into the pT109luc44 vector upstream of the minimal HSV-thymidine kinase (tk) promoter and luciferase (luc) gene. The RARE₂-tkluc reporter vector was previously described.41 Mammalian two-hybrid expression vectors were derived from pGALO and pNLVP16 plasmids45 by subcloning indicated cDNAs in frame with the coding regions for the GAL4 DNA binding and VP16 activating domains, respectively. The GAL(RE)₅-tkluc reporter was derived from the pT109luc plasmid by inserting five copies of the GAL4 DNA binding site upstream of the minimal HSV-tk promoter.

All GST fusion proteins were prepared using standard procedures.40 ³⁵S-methionine–labeled proteins were synthesized in vitro using coupled transcription-translation system, TNT (Promega, Madison, WI), following the supplier's directions. ³⁵S-labeled proteins were incubated with 1 μg GST or a given GST fusion protein (see below for conditions). In the case of ³⁵S-labeled RARα, PML-RARα, and PLZF-RARα, binding reactions were performed for 2 hours at 4°C in the absence or in the presence of 1 × 10⁻⁶ mol/L, 1 × 10⁻⁷ mol/L, or 1 × 10⁻⁸ mol/L ATRA. Assays were performed in NETN buffer (20 mmol/L Tris, pH 8.0, 100 mmol/L NaCl, 1 mmol/L EDTA, 0.5% NP-40) at 4°C for 60 minutes with gentle rocking. Glutathione-Sepharose beads were washed five times with H buffer (20 mmol/L HEPES, pH 7.7, 50 mmol/L KCl, 20% glycerol, 0.1% NP-40). Bound proteins were eluted in Laemmeli loading buffer and separated on a 5% or 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Gels were fixed in 25% isopropanol and 10% acetic acid, dried, and exposed to Kodak Biomax film (Eastman Kodak, Rochester, NY). Anti-Sin3A and anti-N-CoR antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). For coimmunoprecipitation, in vitro translated Sin3A and N-CoR were incubated with a given ³⁵S-labeled protein in NETN buffer for 1 hour at 4°C. Immunocomplexes were isolated by further incubation with an appropriate antibody preadsorbed on protein A/G Sepharose (Pharmacia, Uppsala, Sweden), washed 5 times in H buffer, analyzed by SDS-PAGE, and visualized by autoradiography.

Transient transfections using 293T⁴⁶ or CV-147cells, maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal calf serum (FCS), were performed using either the calcium phosphate precipitation method41 or SuperFect transfection reagent (Qiagen, San Clarita, CA). FCS treated with dextran-coated charcoal was used for all transfection experiments involving subsequent additions of ATRA. Cytomegalovirus (CMV)-driven β-galactosidase gene expression plasmid (CMV-lacZ) was used as a control for transfection efficiency. Cells were harvested for assaying luciferase and β-galactosidase activities 26 to 48 hours after transfection. Luciferase assays were normalized using units of β-galactosidase activity. Both assays were performed using commercial reagents according to the supplier's recommendations (Promega). Cells were treated with either ATRA at 1 × 10⁻⁶ mol/L (Sigma, St Louis, MO) for 20 hours alone or ATRA plus 1 mmol/L sodium butyrate (NaB; Sigma). BSM⁺ DNA (Stratagene) was used as a carrier to equalize the total amount of transfected DNA. In all cotransfection experiments, the total amount of transfected mammalian expression vector was kept constant. All transfections were performed at least three times. The NB-4 cells48 were maintained in RPMI with 10% FCS. Cells were treated either with ATRA alone (500 nmol/L) or with ATRA followed by 1 mmol/L NaB. Cell differentiation was assayed by scoring the percentage of nitro blue tetrazolium (NBT)-positive cells in treated cells versus untreated controls at 2 and 3 days after treatment.

Inhibition and stimulation of granulocytic differentiation by dominant negative RARα mutants49,50 (or antagonists of RARα^51,52) and RARα-specific agonists,52,53respectively, suggest that granulopoiesis requires activation of RARα transcriptional activity by physiologic concentrations of ATRA. The PML-RARα chimera expressed in APL cells inhibits granulocytic differentiation under physiologic (10⁻⁸ mol/L) but not pharmacologic (10⁻⁶ mol/L) concentrations of ATRA,54,55 suggesting lower sensitivity to ATRA. The PLZF-RARα chimera appears to be even less sensitive, or completely insensitive, to ATRA, because APL cells with PLZF-RARα^16,17 or myeloid cells overexpressing transfected PLZF-RARα⁵⁶ fail to respond to ATRA. Nevertheless, the wild-type RARα, PLZF-RARα, and PML-RARα possess approximately equal binding affinities for ATRA.40,57 Given this background information, we attempted to test whether the chimeric receptors differ in their binding affinities for nuclear receptor corepressor N-CoR and, hence, would display lower responsiveness to ATRA. Using an in vitro interaction assay, we found that, in the absence of ATRA radiolabeled RARα, PML-RARα and PLZF-RARα were specifically retained on a matrix-bound fusion of GST with N-CoR fragment containing amino acids 1679-2453 (GST-N-CoR 1679-2453; Fig1B). Furthermore, over ATRA concentrations ranging from 1 × 10⁻⁸ mol/L to 1 × 10⁻⁶ mol/L, binding of RARα, PML-RARα, and PLZF-RARα to GST-N-CoR displayed dramatically different ligand sensitivities. Both RARα and PML-RARα lost most of their N-CoR binding at increasing concentrations of ATRA, although PML-RARα required a higher ATRA concentration to attain a similar degree of dissociation (Fig 1B, compare lanes 2 through 6 between the upper and middle panels). In contrast, binding of N-CoR to PLZF-RARα remained relatively unchanged, even in the presence of pharmacologic concentrations of ATRA (Fig 1B, bottom panel).

The above-noted in vitro data were fully corroborated by results from in vivo experiments that used the mammalian two-hybrid assay with the N-CoR protein fused to the GAL4 DBD and RARα, PLZF-RARα, or PML-RARα VP16 activation domain fusions. As reflected by the luciferase activities generated from the GAL(RE)₅-tkluc reporter vector, strength of in vivo interaction between GAL4(DBD)-N-CoR and VP16-PLZF-RARα was virtually unaffected by 10⁻⁶ mol/L concentration of ATRA (Fig 2). In contrast, when compared with the results obtained in the absence of ATRA, the degree of in vivo association between GAL4(DBD)-N-CoR and VP16-PML-RARα was very low at 10⁻⁶ mol/L ATRA, and GAL4(DBD)-N-CoR/VP16-RARα interaction was nearly completely abolished with ATRA treatment (Fig2).

Differential sensitivities of N-CoR interaction with RARα and the two fusion proteins could be due to stabilization of corepressor binding by either PML or PLZF sequences in the fusion proteins. We therefore used in vitro coimmunoprecipitation and GST-pulldown assays to investigate if the wild-type PML and PLZF could associate with the N-CoR protein. Although PML did not interact with N-CoR, we detected interaction between N-CoR and the PLZF protein (Fig3A). We then used various PLZF and N-CoR deletion mutants to identify domains within the PLZF and N-CoR proteins that were responsible for this interaction (Fig 3B through D). The interaction was mapped to the region of N-CoR that overlapped with the previously mapped Sin3 interaction domain29,30 and the BTB/POZ-domain of the PLZF protein. These results were consistent with the data, described in the preceding section, showing ATRA insensitive interaction between GST-N-CoR (amino acids 1679-2453) and PLZF-RARα (containing amino acids 1-455 of the PLZF protein). Because previous studies have shown that N-CoR corepressor associates with certain nuclear receptors as a complex with Sin3A and HDAC, we determined whether PLZF could also interact with those molecules. Using GST-pulldown assays, we find that in vitro PLZF also appears to interact directly with the Sin3A and Sin3B proteins (Fig 3E) as well as HDAC1 (Fig 3F) and, to a lesser degree, also with HDAC2 (data not shown). Nevertheless, the interaction of the Sin3 proteins with PLZF was considerably weaker than with N-CoR. Interaction of PLZF with HDAC1 was also weak, albeit comparable to the interaction of HDAC1 with the Sin3A protein as detected by coimmunoprecipitation in vitro. All interactions appeared to be specific for PLZF, because no signal was detected using just the GST protein as a control. The in vitro association between PLZF and N-CoR, Sin3, or HDAC1 protein was insensitive to ATRA treatment (data not shown).

The carboxy-terminal Zn-fingers of the PLZF protein possess DNA binding activity with specificity for 5′-GTACAGTAC-3′ motif, which is present in genomic sequences and, serendipitously, in the LexA operator (Sitterlin et al58 and our unpublished results). Furthermore, the N-terminus of the PLZF protein fused to the Gal4 DBD repressed transcription from a Gal4 binding site.59 We now show that the wild-type PLZF protein, which is transiently expressed in 293T cells, represses transcription from its binding site within the LexA operator located upstream of the minimal (−109 to +52) HSV-tk promoter (Fig 4A). The level of repression increases with increasing amounts of cotransfected PLZF expression vector and is completely dependent on the presence of PLZF BTB/POZ-domain. Cotransfecting N-CoR and Sin3A expression vectors further reduces the transcription from the L4-tkluc reporter in the presence but not in the absence of cotransfected PLZF protein (Fig 4B). Treatment of transfected cells with a histone deacetylase inhibitor, such as Trichostatin A (TSA),60 relieved the repression by about 50% (Fig 4C), suggesting that, in addition to histone deacetylation, the PLZF protein represses transcription through other mechanism(s).

In a number of cell systems, the PML-RARα and PLZF-RARα chimeric proteins were found to antagonize the function of endogenous retinoic acid receptors and/or perform as much less potent activators when compared with the wild-type RARα.1,2,8,39-41 When assayed side by side, PLZF-RARα was a consistently poorer activator of transcription from the RARE₂-tkluc reporter than PML-RARα and/or a stronger inhibitor of transcription by the wild-type RARs (Fig 5A). Involvement of histone deacetylation in the mechanism of PML- and PLZF-RARα action is supported by data showing that a known histone deacetylase inhibitor, NaB,61-63 can relieve the repression of ATRA response by these chimeric proteins (Fig 5A). Data obtained using NaB were fully corroborated when another histone deacetylase inhibitor, TSA, was used (not shown). Furthermore, histone deacetylase inhibitors, NaB (Fig 5B) and TSA (data not shown), synergized with ATRA in differentiation of t(15;17)-positive APL cells. It is worth noting that, in the presence of ATRA, other agents, such as hexamethylene-bisacetamide, have been shown to have a rapidly synergistic effect on differentiation of NB-4 cells,64raising possibilities that such compounds may also possess histone deacetylase-inhibiting activities. Combination of ATRA and NaB also synergized to induce differentiation of the PML-RARα–negative human myeloid leukemia HL-60 cell line,65 most likely reflecting stimulatory effects of histone deacetylase inhibitors on the activation of the wild-type RARα by retinoids.

It now appears that the strong inhibitory effect of the PLZF-RARα chimera is based on the ability of its N-terminal BTB/POZ-domain to associate independently, in an ATRA-insensitive manner, with the N-CoR corepressor and possibly also with other components of the nuclear receptor corepressor complex, such as Sin3 and HDAC. Therefore, the PLZF-RARα protein can engage with N-CoR by means of the receptor interaction domain (CoR-box, see Fig 1A) and the PLZF BTB/POZ interaction domain, with the latter being completely insensitive to ATRA. Because PML does not appear to interact with N-CoR, decreased sensitivity of PML-RARα/N-CoR association to ATRA may be due to steric effects that the PML moiety exerts on the N-CoR binding RARα region (CoR-box) of the chimeric protein or some other factors that may interact with both PML and N-CoR and stabilize the complex. In this respect, noteworthy is our previous study reporting interaction and colocalization between PML and PLZF as well as PLZF and PML-RARα proteins (but not PML and PLZF-RARα) in APL cells,18suggesting that PLZF may play a wider role in APL than previously supposed from the study of few patients with the t(11;17)(q23;q21) chromosomal translocation. The hypothesis that PLZF may be such a factor is supported by observation that overexpression of the PLZF protein strongly inhibits responsiveness of the PML-RARα to ATRA as well as ATRA and NaB (Fig 5A). The strong inhibitory effect of PLZF, even in the presence of histone deacetylase inhibitor, is consistent with a model in which transcriptional repression by PLZF is only partially mediated through histone deacetylation. Nevertheless, interaction between PML-RARα and N-CoR remains sensitive to treatment with pharmacologic doses of ATRA providing a mechanism for therapeutic response of APL with t(15;17) to this drug.

Very recently, it has also been shown that PLZF,66 as well as BTB/POZ-domain Zn-finger protein LAZ-3/BCL-6,67 which plays a role in the pathogenesis of diffuse large-cell lymphoma,68,69 interact with another nuclear receptor corepressor, SMRT. In vitro interactions of SMRT with RARα, PML-RARα, and PLZF-RARα also displayed differential sensitivities to ATRA.66 LAZ-3/BCL-6 can also interact with the N-CoR protein (data not shown), suggesting that both nuclear receptors and BTB/POZ-domain Zn-finger proteins repress transcription through the same general mechanism. These results also provide evidence for a relationship between signalling through distinct families of transcriptional regulators, such as nuclear receptors, BTB/POZ domain Zn-finger proteins, and E-box binding Mad/Max proteins. Because distinct regulators often possess common coregulators, by sequestration of a shared coregulator, activation or repression by a given factor may have the opposite effect on activity of another factor(s). For example, activation of a nuclear hormone receptor by ligand binding causes association of a coactivator, such as CBP (CREB-binding protein), which is also required by a number of other activators, including CREB, AP-1, and STAT-1α.70,71 Depending on the relative affinities of these proteins for CBP and their concentrations, activation of a nuclear receptor by its ligand could inhibit CREB, AP-1, and STAT-1α activities and, hence, signalling through cAMP, Ras, and interferon γ (INFγ) pathways, respectively. Because Sin3/N-CoR/HDAC are also required for transcriptional repression by Mad/Max complex,30,36,37 which is associated with cellular differentiation, the fusion receptors may in part contribute to leukemogenesis by sequestration of these factors and inhibition of Mad/Max activity in favor of growth promoting Myc/Max function.

Deregulation of histone acetylation has for some time been thought to be associated with oncogenesis, and histone deacetylase inhibitors were reported to possess anti-tumor activities.72,73 Association of histone acetyltransferase, CBP, with two different leukemogenic translocations, t(11;16)(q23;p13)74 and t(8;16)(p11;p13)75 involving the MLL and MOZ genes, respectively, has also been reported. In summary, our data directly implicate for the first time the nuclear receptor corepressor/histone deacetylase complex in carcinogenesis and provide a plausible mechanism (Fig 6) for the molecular pathogenesis of APL and its response to treatment with ATRA. Central to this model is the observation that nuclear receptor corepressor binding to PML- and PLZF-RARα displays abnormal sensitivity to ATRA, with corepressor/PLZF-RARα association being the least sensitive or not sensitive at all. Because histone deacetylase inhibitors considerably potentiate effects of ATRA on both the activities of the RARα fusion proteins and APL cell differentiation, they may serve as useful adjuncts to ATRA in treatment of APL.

The authors thank Q.-H. Huang, Z. Chen, L. Wiedemann, T. Enver, J. Licht, and M. Greaves for helpful comments and discussions. We are grateful to P. Chambon, S. Hollenberg, C.D. Laherty, R.N. Eisenman, C.A. Hassig, S.L. Schreiber, C.K. Glass, C.V. Dang, E.R. Fearon, T.H. Rabbitts, M. Lanotte, and M. Yoshida for their generous gifts of molecular clones, expression vectors, antibodies, cells, and drugs that were used in this study.