The promyelocytic leukemia–retinoic acid receptor α (PML-RARα) protein of acute promyelocytic leukemia (APL) is oncogenic in vivo. It has been hypothesized that the ability of PML-RARα to inhibit RARα function through PML-dependent aberrant recruitment of histone deacetylases (HDACs) and chromatin remodeling is the key initiating event for leukemogenesis. To elucidate the role of HDAC in this process, we have generated HDAC1–RARα fusion proteins and tested their activity and oncogenicity in vitro and in vivo in transgenic mice (TM). In parallel, we studied the in vivo leukemogenic potential of dominant negative (DN) and truncated RARα mutants, as well as that of PML-RARα mutants that are insensitive to retinoic acid. Surprisingly, although HDAC1-RARα did act as a bona fide DN RARα mutant in cellular in vitro and in cell culture, this fusion protein, as well as other DN RARα mutants, did not cause a block in myeloid differentiation in vivo in TM and were not leukemogenic. Comparative analysis of these TM and of TM/PML−/− and p53−/− compound mutants lends support to a model by which the RARα and PML blockade is necessary, but not sufficient, for leukemogenesis and the PML domain of the fusion protein provides unique functions that are required for leukemia initiation.
Acute promyelocytic leukemia (APL) is characterized by nonrandom reciprocal translocations that always involve the retinoic acid receptor α (RARα) gene on chromosome 17. RARα fuses to the promyelocytic leukemia (PML) gene in the vast majority of APL cases (1, 2). These chromosomal translocations generate X-RARα and RARα-X fusion proteins. X-RARα fusion proteins are oncogenic in vivo (2–6).
APL is characterized by a distinctive block of differentiation at the promyelocytic stage of myeloid development and by unique sensitivity to retinoic acid (RA) treatment (1, 2). RARα binds to retinoic acid response elements (RARE) as a heterodimer with RXRα (1). In the absence of RA, the RARα/RXRα heterodimer inhibits transcription through recruitment of histone deacetylases (HDACs; e.g., HDAC1), nuclear receptor corepressors such as N-CoR or SMRT, and DNA methyltrasferases (DNMT) (7). In the presence of a physiological concentration of RA (10−8 M), the RARα/RXRα heterodimer dissociates from the HDAC complex and recruits transcriptional coactivators (8). In contrast, at physiological RA concentration, PML-RARα protein acts as a dominant negative (DN) RARα by forming aberrant complexes with HDAC and DNMT through the PML moiety of the fusion protein (4, 8–11). At a pharmacological dose of RA, PML-RARα releases the HDAC complex and activates transcription, thus mimicking RARα. Point mutations have been reported in the RARα ligand-binding domain of PML-RARα in cases with acquired resistance to RA (12). Collectively, these data suggest that aberrant recruitment of HDAC to RARE represents a key event in APL leukemogenesis. However, the PML-RARα oncoprotein can also interfere with the function of the remaining PML allele through heterodimerization (1, 2), and it remains to be determined to what extent each of these processes contributes to APL leukemogenesis.
Results And Discussion
To determine whether aberrant HDAC-dependent transcriptional repression is necessary and sufficient for leukemogenesis, we generated transgenic mice harboring the following: (a) DN RARα mutants along with their PML-RARα counterpart and (b) an artificial HDAC–RARα fusion protein along with its enzymatically inactive counterpart. We also studied in vivo an RARα truncated mutant corresponding to the moiety of RARα invariably shared by all the APL fusion proteins (1, 2) (Fig. 1 A
RARαE carries a glycine (G) to glutamate (E) substitution at amino acid 303 in the RARαE domain that is responsible for ligand binding. This mutation leads to RA resistance and in vivo photocopies the RARα KO phenotype (13). RARαM4 carries a leucine (L) to proline (P) substitution at amino acid 398 in domain E; and PML-RARαM4 harbors the same mutation found in RARαM4 (14). This mutation leads to RA-insensitive transcriptional repression (14).
HDAC1-RARα expresses the full-length HDAC1 coding sequence fused to RARα. HDAC1 is part of the aberrant PML-RARα transcription (4, 9, 10). mHDAC1-RARα carries a point mutation that abrogates HDAC1 enzymatic activity (histidine to phenylalanine at HDAC1 amino acid 199) (15). ΔRARα carries a deletion that removes domain A from RARα. This deletion is identical to the one observed in the X-RARα fusion proteins and removes a domain responsible for transcriptional activation function (1, 16). These constructs were cloned in the human cathepsin-G (hCG) minigene (3, 4) and used to generate transgenic lines (Fig. 1, B and C).
We assessed whether HDAC1-RARα displayed the distinctive features of the X-RARα fusion proteins. We found that HDAC1-RARα can homodimerize and heterodimerize with RXRα within the cell (Fig. 2, A and B
). HDAC1-RARα can effectively bind to the DR5 consensus sequence. Electromobility shift analysis (EMSA) produced a single HDAC1–RARα protein DNA complex, whereas HDAC1-RARα with RXRα formed two complexes (Fig. 2 C). These bands were abolished by the addition of an excess of unlabeled DR5 and super shifted with specific antibodies (Fig. 2 C). These data demonstrate that HDAC1-RARα forms homo- and, and more efficiently, heterodimers that are able to bind to the DR5 consensus sequence, as previously demonstrated in the case of other APL fusion proteins (17, 18).
Next, we investigated whether HDAC1-RARα acts as a transcriptional repressor. Vectors expressing RARα, PML-RARα, PLZF-RARα, HDAC1-RARα, mHDAC1-RARα, and HDAC1 were transfected into 293T cells together with RARβ-luc, a luciferase reporter construct containing the RARα-responsive promoter region of RARβ. Luciferase assays demonstrated that HDAC1-RARα acted as a potent transcriptional repressor (Fig. 3 A
). As expected as the result of disruption of HDAC1 enzymatic activity, mHDAC1-RARα showed a much weaker transcriptional repression. HDAC1-RARα, PLZF-RARα, and PML-RARα repressed transcription equally well in the presence of RA, whereas mHDAC1-RARα did not (Fig. 3 A). HDAC1-RARα, therefore, acts as an aberrant transcriptional repressor and this property depends on the HDAC1 enzymatic activity.
Chromatin immunoprecipitation (ChIP) experiments on the promoter of the cytoplasmic retinoic acid binding protein II (CRABPII) gene revealed that HDAC1-RARα inhibited acetylation of histone H3 (Fig. 3 B). HDAC1 and HDAC1-RARα both inhibited histone H3 and H4 acetylation by the bromodomain of the p300 protein (19). This inhibition was partially abrogated with mHDAC1-RARα (Fig. 3 C). Thus, HDAC1-RARα displays HDAC activity.
Because both PML-RARα and PLZF-RARα block TGFβ1 and vitamin D3–induced cellular differentiation of U937 cells (20, 21), we tested whether constitutive expression of HDAC1-RARα affected cellular differentiation upon TGFβ1 and vitamin D3 treatment. We found a significant reduction in the induction of the myeloid marker CD11b in cells transduced with MIGR1-PLZF-RARα (P = 0.01, calculated by the Student's t test) and MIGR1-HDAC1-RARα (P = 0.02, calculated by the Student's t test), whereas MIGR1-mHDAC1-RARa and MIGR1-HDAC1 exerted no significant effect on myeloid differentiation (Fig. 3 D). Collectively, these data suggest that HDAC1-RARα shares many of the features of the X-RARα protein, including its ability to act as a transcriptional repressor of RARα through HDAC activity.
) (3, 4). The transgene was invariably expressed (Fig. 1 C). Leukemia was observed in three PML-RARαM4 transgenic lines. Latency was 8–9 mo (Fig. 4 A), in agreement with what we observed in PML-RARα transgenic lines (3). Strikingly, only 1 of the RARαE transgenic lines out of the 19 lines expressing DN RARα mutants (ΔRARα, RARαM4, RARαE, and HDAC1-RARα) developed leukemia after a long latency (18–19 mo) and at low penetrance (Fig. 4, A–C). Morphological analysis of the leukemic bone marrows and spleens revealed the presence of blasts with promyelocytic features. Flow cytometric analysis with Mac-1, Gr-1, c-kit, B220, CD3, and Ter119 cell surface markers of the confirmed the diagnosis of APL (Fig. 4, B and C, and not depicted). RARaE-induced leukemias were transplantable in secondary recipients and leukemic mice showed no response to RA treatment as compared with PML-RARα leukemic mice (Fig. 4 D) (mean survival time: 10.4 d; 95% confidence interval = 1.9–18.9 d vs. mean survival time: 44.3 d; 95% confidence interval = 36.7–51.9 d) (22).
The RARα gene is invariably involved in the APL-associated chromosomal translocations (1, 2). Therefore, alteration of RARα pathway has been thought to play a central role in APL pathogenesis. Indeed, RA inhibits the proliferation of hematopoietic precursors and promotes the terminal granulocytic differentiation of granulocyte/monocyte progenitors and multipotent erythroid/monocytic cells. Vitamin A deficiency, unligated RARα, RARα antagonist, or DN RARα can block myeloid differentiation (23). Moreover, the X-RARα fusion proteins can block differentiation when overexpressed in myeloid leukemia cell lines such as U937 cells and interference with PML function seems not to be required for this function (20, 21). These observations support the notion that DN blockade of the RARα pathway is crucial for APL leukemogenesis. Our in vivo genetic analysis challenges this notion, allowing us to reach three major conclusions.
The first major conclusion is that HDAC1-dependent DN blockade of RARα function is neither sufficient to cause leukemia nor to block myeloid differentiation in vivo. The fact that only PML-RARα and PML-RARαM4 (which retain the X moiety), but none of the other DN RARα mutants triggered leukemia in multiple transgenic lines demonstrates that inhibition of RARα per se is not sufficient to initiate leukemogenesis. Our experiments do not rule out that HDAC-chimeric constructs other than HDAC1-RARα may display a leukemogenic effect. Indeed, corepressors do not solely recruit HDAC1, but also other types/classes of histone deacetylases, and PML interacts with both HDAC1 and 2. However, prior observations support this conclusion as PLZF-RARα, transgenic mice develop leukemia, but not a block of myeloid differentiation, whereas RARα−/− mice display a normal myeloid differentiation (4, 24).
The second major conclusion is that only one out of the six RARαE lines developed APL after a long latency (1.5 yr) with very low incidence. This observation strongly suggests that blockade of RARα function is necessary, but not sufficient, for leukemogenesis. Interestingly, these leukemias were resistant to RA, demonstrating that RARαE functions in these leukemic cells as an RA-insensitive receptor.
The third major conclusion is that PML moiety is important in leukemogenesis not solely because it permits aberrant recruitment of HDAC1 and HDAC2, DNMT or homodimerization (11, 18, 25, 26) but also because it interferes with the tumor suppressive function of the wild-type PML gene product. Indeed, only PML-RARα and PML-RARαM4 lead to a DN disruption of the PML-NB both in vitro and in vivo (unpublished data). The critical role of PML functional inactivation is further underscored by the fact that APL is dramatically accelerated in PML-RARα/PML−/− mice (27). In addition, through the PML/X moiety, the fusion protein acquires aberrant gain-of-function properties (e.g., aberrant DNA binding activity) (28). Indeed, it has been shown that PML-RARα homodimer binds specific DNA sites that are not preferentially recognized by the RARα/RXRα heterodimer, thus suggesting the possibility that X-RARα may exert oncogenic functions that are not derived from its DN activity against the RARα/RXRα heterodimer (28, 29). This is supported by the fact that neither RARαM4 nor RARαE triggered leukemia even in the absence of PML: MRP8-RARαM4/PML−/− and MRP8-RARαE/PML−/− mice did not develop leukemia during a 12-mo follow up (unpublished data and Kogan, S., personal communication). Interestingly, RARαM4 did not trigger leukemia in the absence of p53, either; MRP8-RARαM4/p53−/− compound mutants succumbed to lymphoma with incidence and onset similar to p53 null mice (Kogan, S., personal communication).
We propose a model by which the combined inactivation of the X and RARα pathways are both required, but not sufficient, for tumor initiation. PML-RARα is bestowed with additional PML-dependent functional gains that critically contribute toward full-blown transformation. Nevertheless, additional genetic abnormalities are required for leukemogenesis even in the presence of the full-length oncogenic fusion protein, as strongly suggested by the long leukemia latency observed in any of the X-RARα transgenic models and the recurrent chromosomal abnormalities that the leukemic blasts from these models invariably harbor at presentation (30, 31).
On the basis of this model, it remains to be explained why RA and HDAC inhibitors are effective in APL treatment. In this respect, it is tempting to speculate that the blockade of the RARα pathway, while not sufficient for leukemia initiation, may be necessary for leukemia maintenance.
Materials And Methods
Cells and expression vectors.
Cells were obtained from the American Type Culture Collection. Vitamin D3 was obtained from Sigma-Aldrich and TGFβ1 was obtained from PeproTech. Plasmids expressing RARα, RXRα, RARαM4, PML-RARα, PML-RARαM4, PLZF-RARα, HDAC1-FLAG (provided by P.A. Marks and V. Richon, Memorial Sloan Kettering Cancer Center, New York, NY), His-BrHAT (provided by A. Tomita, Nagoya University, Nagoya Aichi, Japan), and RARαE have been described previously (4, 13, 14, 19, 32). pSG5-HDAC1-RARα carries the full-length HDAC1 gene fused in frame with the full-length RARα. Mutant HDAC1-RARα (mHDAC1-RARα) was generated by site-directed mutagenesis. pSG5-ΔRARα was generated by PCR. pCMV-PML-RARα, pCMV-HDAC1-RARα, pCMV-mHDAC1- RARα, and pCMV-HDAC1 are pCMV-Tag 2B (FLAG-tagged) derivative (Stratagene). pCDNA3.1/His-HDAC1-RARα, pCDNA3.1/His-RXR, pCDNA3.1/His-RARαE, and pCDNA3.1/His-RARαM4 were obtained by cloning the respective cDNAs into pCDNA3.1/His C (Invitrogen). To generate retroviral constructs, Flag-tagged RARα, PLZF-RARα, HDAC1-RARα, mHDAC1-RARα, and HDAC1 were cloned into pMIGR1. The sequence of each vector was confirmed sequencing.
Antibodies, immunoprecipitations, and Western blot analyses.
We used the antibodies specific for: RARα (C-20) and RXRα (D-20) (Santa Cruz Biotechnology, Inc.), histone H3, H4, acetylated histone H3 and H4 (Upstate Biotechnology); PML (Chemicon International), M2 anti-Flag (Sigma-Aldrich), and anti-Xpress (Invitrogen).
Gel shift assay.
RARα, RXRα, and HDAC1-RARα proteins were generated in vitro by TNT Coupled Reticulocyte Lysate Systems (Promega). Protein synthesis was confirmed by Western blot. Aliquots were used for gel shift analysis with the 32P-labeled DR5 oligonucleotide: 5′-GGGACAAAGGTCAACGAAAGGTCAGAGCCC-3′ (29). For competition assays, we used 100-fold molar excess of unlabeled DR5. For supershift experiments, we used anti-RARα, anti-RXRα antibodies, or normal rabbit IgG (Santa Cruz Biotechnology, Inc.).
293T cells were cotransfected with RARβ-luc and pRL-TK (encoding firefly and renilla luciferases, respectively) and the relevant pSG5 expression constructs using Effectene Transfection Reagent (QIAGEN). Cultures were treated with 10−6 M of RA 24 h after transfection. Luciferase and renilla assays were done 48 h after transfection.
Chromatin immunoprecipitation (ChIP) assay.
We used the ChIP assay kit (Upstate Biotechnology).
Retroviral transduction and flow cytometry analysis of U937 cells.
Recombinant retroviruses were used to transduce U937 cells by spinoculation for three consecutive days. GFP-positive cells were sorted with MoFlo (DakoCytomation). Expression of mutant RARα was confirmed by Western blot. CD11b was quantified by FACScan (BD Biosciences). These experiments were repeated five times. The unpaired Student's t test was used to compare CD11b expression between cells transduced with the MIGR1 vector and the ones transduced with MIGR1 vectors expressing PLZF-RARα, PML-RARα, HDAC1-RARα, and mHDAC1-RARα.
Southern blot analysis.
Total RNA was extracted from mouse bone marrow with TRIzol (Invitrogen) and treated with DNase I. RT was performed using 2 μg of total RNA with SuperScript First-Strand Synthesis System (Invitrogen). 1 μl of cDNA was used for nested PCR.
Follow up of transgenic mice.
Bone marrow transplants in nude mice and ATRA treatment.
Leukemic cells were obtained from bone marrow and spleens of leukemic RARαE TM. 4–8-wk-old Nu/J Hfh 11nu nude mice were injected with 2 × 106 leukemic cells intravenously. Transplanted nude mice (NM) were bled once a week. The leukemic TM and the NM that developed leukemia after transplantation received intraperitoneal injections of 1.5 μg/g of RA daily (22).
We thank Dr. S. Kogan for sharing unpublished observations and for critical discussions.
This work was supported by National Institutes of Health grant RO1 CA-74031 (to P.P. Pandolfi). H. Matsushita received a postdoctoral Fellowship from the Uehara Memorial Foundation and P.P. Scaglioni was supported by the American Society of Clinical Oncology Young Investigator Award, the CALGB Oncology Fellows Award, the Charles A. Dana Foundation, the Michael and Ethel L. Cohen Foundation, and the Steps for Breath Foundation. W.H. Miller Jr. is supported by the Canadian Institutes of Health Research and is a Chercheur National of the Fonds de la Recherche en Santé du Québec.
The authors have no conflicting financial interests.