INTRODUCTION

Acute promyelocytic leukemia (APL) represents approximately 10% of all adult acute myeloid leukemias (1). The molecular pathogenesis of APL is, at least in part, associated with the disruption of the retinoic acid receptor  (RAR) gene through its fusion to one of four different loci (2-7). These translocations result in the expression of chimeric RAR fusion proteins that retain the DNA and ligand binding domains of the receptor and gain a dimerization domain from the fusion partner. Paradoxically, APL is the first human malignancy that may undergo complete remission in response to differentiation therapy with all-trans-retinoic acid (RA). The molecular basis of these remissions is still disputed.

The majority of APL cases, and all cases that consistently respond to RA treatment, possess the t(15;17) translocation that fuses the PML and RAR genes (3, 8-12). PML is a member of a functionally diverse gene family that encodes proteins characterized by the presence of a N-terminal C₃HC₄ RING-finger motif (13), followed by one or two cysteine-rich regions (B boxes) and a coiled-coil protein dimerization interface. The function of PML is unknown, but up-regulation of its expression by interferons (IFNs) (14-16) and its negative effect on cell growth and cellular transformation by cooperating oncogenes (17-19) suggest a role in growth control. The product of the wild-type PML gene is a phosphoprotein (20) that localizes both in the nucleoplasm and in the specific multiprotein structures called PML nuclear bodies (NBs) (20-24). The PML/RAR fusion protein, which is expressed in APL cells as a result of t(15;17), contains all predicted PML structural motifs and is able to delocalize the wild-type PML and other NB components onto discrete microspeckled nuclear structures (21-24). It is still unclear which role, if any, disruption of NBs and/or establishment of microspeckled structures play in cellular transformation. Nevertheless, complete restoration of NBs upon RA treatment in NB4 cells, but not in RA-resistant sublines (which remain microspeckled), strongly suggests a perturbation of NBs in the pathogenesis of APL (23, 25).

Three other APL-associated translocations of the RAR gene have been characterized at the molecular level. The t(5;17)(q35;q21) (5), t(11;17)(q23;q21) (6), and t(11;17)(q13;q21) (7) fuse RAR to nucleophosmin (NPM), promyelocytic leukemia zinc (Zn) finger (PLZF), and nuclear mitotic apparatus (NuMA) genes, respectively. In all cases RAR fusion proteins are expressed that structurally resemble PML/RAR. All four fusion proteins possess identical RAR sequences, which include the DNA binding region C, the ligand binding domain E, and N-terminal protein-protein interaction motifs derived from PML, PLZF, NuMA, and NPM proteins. Nevertheless, at least in the case of PLZF/RAR-associated APL, patients and primary PLZF/RAR-positive APL cells in culture do not respond to RA (26). The molecular basis for this difference in clinical response of the disease is not understood.

Recently, we have shown that in analogy to PML, the PLZF protein possesses a speckled nuclear localization (27, 28) and its overexpression leads to suppression of cell growth (A.R., Rita Shaknovich, J.L., S.W., and A.Z., unpublished data). PLZF belongs to a protein family characterized by the presence of a BTB/POZ domain, involved in dimerization and transcriptional repression (29, 30). Moreover, most of these proteins contain several Krüppel-like Zn fingers, some of which are involved in sequence-specific DNA binding (29, 31). PLZF/RAR contains the N-terminal BTB/POZ domain and two out of nine Krüppel-like Zn fingers fused upstream of the RAR sequences. When transiently expressed in mammalian cells, this protein also appears to be, at least in part, localized to discrete nuclear domains (32).

The similarity in the intranuclear localization of PML and PLZF, as well as their RAR fusions, prompted us to investigate whether PML and PLZF are present in the same nuclear compartments and might be functionally related. Herein we provide evidence for an interaction and nuclear colocalization between these two proteins. Apart from identifying a new PML partner, these findings strengthen the importance of NB disturbances in APL and unify the molecular models for its pathogenesis.

MATERIALS AND METHODS

Transfections of COS6 cells with 0.5 µg of PLZF^flag (27) and/or 0.5 µg of PML (8) mammalian expression vectors were carried out in 60-mm tissue culture plates (see below). The total amount of transfected expression vector was kept constant at 1 µg in all experiments. Approximately 18 h after transfection cells were labeled for 4 h with [³⁵S]methionine/cysteine (200 µCi; 1 Ci = 37 GBq), then lysed with 800 µl of lysis buffer [250 mM NaCl/10 mM Tris·HCl, pH 7.4/1 mM EDTA/0.1 mM Na₃VO₄/0.5% Nonidet P-40/2% phenylmethylsulfonyl fluoride/arrotonin (75 µg/ml)/soybean trypsin inhibitor (75 µg/ml)/leupeptin (75 µg/ml)/bestatin (75 µg/ml)], and 700 µl of the cell lysate was incubated at 4°C for 2 h with 2 µg of anti-FLAG antibody (Kodak) [preadsorbed for 2 h at 4°C on 50% slurry of protein G-Sepharose beads (Pharmacia) in lysis buffer]. Proteins were eluted from the beads with Laemmli-loading buffer and electrophoresed on SDS/PAGE gels. Resolved proteins were transferred to a solid support and blotted with an anti-PML mouse monoclonal antibody (see below).

Immunofluorescence was performed essentially as described (21, 28). Polyclonal rabbit anti-PML (21) and the monoclonal mouse anti-PLZF (Rita Shaknovich, J.L., S.W., and A.Z., unpublished results) antibodies were used at a 1:400 and 1:100 dilutions in Tris-buffered saline (TBS), respectively. Confocal microscopy was as described (21).

Transient and semistable transfections were performed in COS6 or CHO cells by using Lipofectin (Life Technologies) according to the manufacturer's instructions. In short, cells were plated at 50% or 30% (semistable) confluence in a 25-cm² flask and transfected in Optimem medium with 3 µg of plasmid overnight. The next day serum was added, and 48 h (transient) or 5-6 days (semistable) later, cells were examined for expression. COS6 and CHO cells were cultured in DMEM with 10% fetal calf serum, and KG1 and NB4 cells were cultured in RPMI 1640 medium with 10% retal claf serum. For IFN treatment, cells were incubated for 2 days with IFN- or  at 1000 units/ml. For RA treatment, 1 day after transfection PLZF/RAR-transfected cells were split 1:2, treated or not with 10⁶ M RA, and harvested 24-48 h later.

GST-PLZF protein was prepared as described (32). For "pull-down" experiments, 1 µg of protein on glutathione-Sepharose beads (Pharmacia) and 3 µl of a given ³⁵S-labeled in vitro-translated product made with TNT-coupled reticulocyte lysate (Promega) were used. All conditions were as described (32).

The Y190 strain of Saccharomyces cerevisiae (33), which harbors integrated lacZ reporter gene under control of the Gal4-responsive promoter, was transformed by using the lithium acetate method (34) with 1 µg of yeast expression vectors, pASI and pACTII (35), containing either PLZF or PML cDNAs fused in-frame to the Gal4 DNA binding and activation domains, respectively. Liquid and solid media, as well as conditions used for yeast growth, were as described (36). For -galactosidase assays, randomly picked colonies were grown in a liquid culture to an OD₆₀₀ of approximately 1.0. Yeast cells were lysed and assayed for -galactosidase activity by using standard procedures (37). Units of -galactosidase activity were calculated by the formula 1000 × (OD₄₂₀/t × V × OD₆₀₀), where t is incubation time in min and V is the volume of extract used for the assay in ml.

Western blot analysis was performed as described (21). The polyclonal anti-PLZF (27, 38) and monoclonal anti-PML (M.-C. Guillemin and H. de Thé, unpublished results) antibodies were used at 1:500 and 1:200 dilutions in TBS, respectively. Anti-p68 kinase antibody (Ribogene, Hayward, CA) was used at 1:10,000 dilution.

RESULTS

Given the parallels between the PLZF and PML proteins, we investigated the localization of PLZF and PML in a number of different systems. In the nuclei of KG1 cells, various degrees of colocalization between endogenous PLZF and PML proteins were observed, ranging from a small number of doubly positive NBs (Fig. 1D) to total colocalization (Fig. 1 A-C). Similar results were found in K562 or HL60 cells (data not shown). Subsequently, we tried to reproduce these results in cotransfection experiments. In COS6 cells, transiently transfected with PLZF and PML expression vectors, complete colocalization was rarely observed; nevertheless in each experiment, cells could be found in which the two proteins totally colocalized (Fig. 1E). However, in CHO cells semistably transfected with both genes colocalization was constantly found (data not shown), suggestive of a time/expression-dependent targeting process.

As most of the PML protein is associated with the nuclear matrix (39) and thus requires stringent conditions for extraction that destroy noncovalent interactions, coimmunoprecipitation of the endogenous PML and PLZF proteins proved technically very difficult. Nevertheless, we were able to readily coimmunoprecipitate PLZF and PML proteins from nuclear extracts of transiently cotransfected COS6 cells (Fig. 2A). This suggested that at least in this artificial system, a fraction of the two proteins are engaged in the same complex.

The interaction is apparently direct because we were also able to show specific binding between in vitro-translated ³⁵S-labeled PML, as well as PML/RAR, with bacterially produced GST-PLZF (summarized in Fig. 2B). A number of in vitro-translated PML mutants and a different C-terminal isoform of PML [herein referred to as PML(B) (10)] also interacted to a similar degree with the PLZF protein in vitro (Fig. 2B). These data implicate the coiled-coil region of PML in the interaction with PLZF. Note, that this same region has also been shown to play a role in PML homodimerization (40).

The PLZF-PML interaction was also assessed in vivo by using the yeast two-hybrid system. As can been seen in Fig. 3, PLZF [fused to the DNA binding domain of Gal4] interacted with the wild-type PML (fused to the Gal4 activation domain). As previously shown, PML and PLZF formed PML-PML and PLZF-PLZF homodimers (32, 40). The C-terminally truncated PML, as well as PML(B), also interacted with PLZF in this assay, corroborating the results obtained from in vitro experiments. However, the interaction between PML and PLZF was unidirectional because it only could be detected when PLZF was fused to the DNA binding domain and PML to the activation domain of Gal4 (data not shown). The directionality of interaction in the two-hybrid system is not unique to this case because it has been demonstrated for other proteins including Myc and Max (41).

Localization of PLZF/RAR and PML/RAR Fusion Proteins.

Previously, we and others had shown that the PML/RAR fusion protein exhibits a dominant negative behavior upon the localization of endogenous PML and, so far, on all the other NB-associated antigens (10, 21-24, 42). Therefore, it was interesting to determine whether the nuclear localization of PLZF would also be affected by the PML/RAR fusion protein. In the t(15;17)-positive NB4 APL cell line, PML/RAR delocalize the endogenous PLZF proteins onto the microspeckles (Fig. 4 C and D). After RA treatment of NB4 cells, the PML pattern is restored and coincides exactly with the normal PLZF distribution (data not shown). These results are readily reproducible in transient or semistable transfections (data not shown) and are consistent with the requirement of the PML coiled-coil domain for in vitro interaction with PLZF. As suggested (32), also PLZF/RAR displays a different fluorescence pattern when compared with the wild-type PLZF (Figs. 4 B, B1, and B2 and 5A1); in both transient or semistable transfections, the fusion protein's pattern is microspeckled. The differences in PLZF/RAR staining patterns among various cells examined within the same transfection experiment (Fig. 4, B-B2) were probably due to the varying levels of PLZF/RAR expression and/or to a given cell's stage in the cell cycle.

6

6

In cells transfected with a PLZF/RAR expression vector (Fig. 4 A-A2) or in primary APL blasts from a patient with a t(11;17)(q23;q21) translocation (data not shown), normal nuclear localization patterns of the endogenous PML were observed. This strongly suggests that at least some of the seven C-terminal Zn fingers of PLZF are required for its interaction with PML and that the PLZF/RAR fusion protein does not affect the PML localization.

As the microspeckled patterns of both PML/RAR and PLZF/RAR resembled each other, we transiently cotransfected vectors encoding each of these fusion proteins. As demonstrated in Fig. 1 F and G, both proteins appear to colocalize to the same nuclear domains. Remarkably, in sharp contrast to the PLZF-PML colocalization, which was rare under transient but common under semistable expression conditions, this colocalization was very frequently observed in transient systems. Furthermore, a PLZF/RAR mutant that lacks the BTB/POZ domain colocalized with the PML/RAR protein, whereas a mutant that lacks the BTB/POZ domain and the two N-terminal Krüppel-like Zn fingers of the PLZF protein or just the two Zn-finger motifs did not (data not shown). These results are in agreement with in vitro binding data of various PLZF deletion mutants to GST-PML that also indicate the requirement of the PLZF Zn fingers, but not the BTB/POZ-domain, for interaction with PML (data not shown). The two Krüppel-like Zn fingers in the PLZF/RAR fusion protein were also required for its full dominant negative effect on the wild-type RAR activity (32).

If PLZF/RAR and PML/RAR colocalization involves a physical contact between these proteins, the above results could suggests that the two N-terminal Zn fingers also participate in the PML-PLZF interaction. PLZF/RAR expression has no effect on localization of the endogenous PML, suggesting that some of the C-terminal PLZF Zn fingers may also be necessary for interaction with the wild-type PML. Therefore, we propose three nonexclusive hypotheses to explain why PML/RAR and PLZF/RAR can interact and/or colocalize: (i) conformational changes that allow an interaction exclusively between the first two Zn fingers, (ii) interaction via the two N-terminal Zn fingers stabilized by the strong DNA binding of the RAR DNA binding domains, and (iii) alternatively, the microspeckled pattern could be explained by a transport of fusion proteins to the same microspeckles as homodimers PLZF/RAR and PLZF/RAR or PML/RAR and PML/RAR but not as heterodimers PLZF/RAR and PML/RAR.

RA Causes Degradation of PLZF/RAR.

We also investigated the effect of RA treatment on localization of PLZF/RAR in transfected cells. COS6 cells transiently expressing PLZF/RAR (or semistable CHO cells), when treated with RA, displayed decreased staining with anti-PLZF antibody (Fig. 5 A1 and A2) and a lower number of PLZF/RAR-positive cells (data not shown). Western blot analysis of extracts derived from treated and untreated cells indicated sharply decreased levels of PLZF/RAR protein (Fig. 5B). Although we cannot exclude the possibility that RA causes elimination of cells expressing PLZF/RAR or inhibits its expression, it is most likely that the PLZF/RAR protein is degraded in RA-treated cells. This conclusion is further supported by marked reduction of PLZF/RAR levels in t(11;17)-positive APL cells after RA treatment (data not shown). These findings are remarkably similar to the recently published results regarding the PML/RAR protein (43-45). It is paradoxical, however, that both fusion proteins are degraded upon RA treatment but that only the t(15;17)-associated APL can be treated by in vivo RA administration (26).

Given the inducibility of PML by IFNs, we investigated whether the expression of the PLZF gene is regulated by these signaling molecules. Although IFN- and IFN- substantially increased both the number and intensity of PLZF-positive NBs (Fig. 5 C1-C3), Western blot analysis revealed an increase in the levels of the IFN-induced p68 kinase but not of the PLZF protein (Fig. 5 D1 and D2). The increased PLZF fluorescence in the NBs is most likely due to the recruitment of nucleoplasmic PLZF onto the NBs by increasing amounts of PML after IFN treatment. Given our recent demonstration that arsenic aggregates PML onto "NBs," the behavior of PLZF after arsenic exposure was investigated. As previously shown for NB-associated proteins (45), together with arsenic, IFN leads to a dramatic aggregation of PLZF and PML onto the bodies, but alone has little or no effect (data not shown).

DISCUSSION

RAR is one of a number of genes that are frequently targeted by chromosomal translocations encountered among different hemopoietic malignancies that are associated with expression of fusion gene products (46). For example, more than 10 different translocations involving MLL gene have been described (47). As with the majority of MLL rearrangements, genes that are fused with the RAR appear to have very little in common except that they all provide RAR with a heterologous dimerization interface, allowing for formation of stable homodimers of the fusion proteins, and may all be involved in some aspects of cell growth control (see above). It is not clear at the present time whether the non-RAR sequences of various chimeric proteins play an active or passive role(s) in transformation. Retinoids and retinoid receptors affect not only differentiation of hemopoietic cells but also cell growth and apoptosis (48, 49). One could therefore argue that the entire APL phenotype, block of differentiation and inhibition of programmed cell death, is a result of a disruption in retinoid signaling by the RAR chimeric proteins.

The first evidence against this hypothesis stems from the fact that in several cellular systems PML is a potent growth suppressor (17-19), and the second stems from the ability of the PML/RAR chimeric protein to delocalize different components of NBs onto microspeckled nuclear structures, a phenomenon reversible in RA-sensitive APL cells upon their differentiation in response to RA (21-24, 50). Moreover, in RA-resistant NB4 mutants that only differentiate in the presence of cAMP, RA followed by cAMP administration, but not RA alone, restores both NB structure and differentiation (25). Furthermore, the dramatic role of arsenic trioxide, which targets PML and PML/RAR onto NBs and induces apoptosis in APL cells (45), strongly suggests a contribution of PML and NB disruption in the pathogenesis of this disease.

Our experiments identify, to our knowledge, the first PML-associated protein for which physical interaction was actually demonstrated. This is in contrast with the recent work on the Pic-1 protein (51) where coimmunoprecipitation and GST "pull down" were not feasible, questioning a non-enzyme-mediated direct interaction with PML. Colocalization and interaction between PML and PLZF strongly suggests a functional relationship for these two proteins. In the context of their respective RAR chimeras, PML and PLZF may also affect the same nuclear processes. This is strongly supported by the fact that both PML/RAR and PLZF/RAR localize in the microspeckled nuclear structures. Some of these microspeckles are sites of transcription (52) and could be important targets for the transforming activities of the fusions. Our studies of PLZF protein indicate that it functions as a DNA binding transcriptional repressor (unpublished data) and that the BTB/POZ domain is required for the repressing effect. In this respect, it is interesting that a Krüppel-like Zn-finger protein with a related transcriptional repression domain, KRAB, interacts with a PML-related RING-finger protein, TIF1 (53), and this interaction is required for transcriptional repression by KRAB proteins (54-56). Note, in this respect that a PML family member, Rpt-1, is a transcriptional repressor of several promoters (57) and that PML appears to enhance the transcription by the progesterone receptor (58). In an analogous manner, PML could participate in transcriptional regulation by the PLZF protein. Alternatively, as suggested by the sharp increase in NB-associated PLZF fluorescence after IFN treatment (and furthermore arsenic), PML could shuttle bound proteins, such as PLZF and SP100 (45), from the nucleoplasm onto the NBs and in this way modulate transcriptional regulation by the PLZF protein.

Interestingly, in NB4 cells PML/RAR delocalizes the wild-type PLZF, whereas PLZF/RAR has no effect on the localization of the wild-type PML, pointing to the importance of alterations in PLZF nuclear localization. In this respect, it is worth noting that the pattern of PLZF expression during hemopoiesis suggests that its product may play a role in myeloid development (28) and perturbation of this potential role by the PML/RAR chimera may contribute to leukemogenesis. Whereas all the data presented herein for PLZF/RAR are derived from studies using transfected cells, results obtained from recent analyses of cells derived from a t(11;17)-positive APL patient (to be reported elsewhere) completely corroborate conclusions drawn in this manuscript. It is remarkable that the two different RAR gene translocations lead to the expression of fusion proteins that colocalize in the same microspeckled nuclear structures and that the common feature of the leukemias involving either PML [t(15;17)] or PLZF [t(11;17)] is the presence of PLZF and RAR sequences in these microspeckles. Although several aspects deserve additional analysis, our findings that PLZF and PML (which both have growth suppressive/apoptotic properties) interact and colocalize in the nucleus, unify not only the cellular targets of the two fusions but also the pathophysiology of APL.