Promyelocytic leukemia (PML) has been implicated in a variety of functions, including control of TP53 function and modulation of cellular senescence. Sumolated PML is the organizer of mature PML bodies, recruiting a variety of proteins onto these nuclear domains. The PML gene is predicted to encode a variety of protein isoforms. Overexpression of only one of them, PML-IV, promotes senescence in human diploid fibroblasts, whereas PML-III was proposed to specifically interact with the centrosome. We show that all PML isoform proteins are expressed in cell lines or primary cells. Unexpectedly, we found that PML-III, PML-IV, and PML-V are quantitatively minor isoforms compared with PML-I/II and could not confirm the centrosomal targeting of PML-III. Stable expression of each isoform, in a pml-null background, yields distinct subcellular localization patterns, suggesting that, like in other RBCC/TRIM proteins, the COOH-terminal domains of PML are involved in interactions with specific cellular components. Only the isoform-specific sequences of PML-I and PML-V are highly conserved between man and mouse. That PML-I contains all conserved exons and is more abundantly expressed than PML-IV suggests that it is a critical contributor to PML function(s). (Cancer Res 2006; 66(12): 6192-8)
- nuclear organization
The human gene promyelocytic leukemia (PML) was identified as the fusion partner of the retinoic acid receptor α (RARA) gene by the t(15;17) chromosomal translocation found in the promyelocytic leukemia (APL; refs. 1, 2). PML is composed of nine exons dispersed on 35 kb ( 3). Exons 6 to 9 could splice alternatively, yielding a large number of isoforms (Supplementary Fig. S1; ref. 4). Expression of the PML gene is sharply induced by IFN ( 5) or TP53 activation ( 6). 4 Tissue-specific expression of PML was described in the myeloid lineage and in endothelial cells ( 7, 8). Inflammation or oncogenic transformation also greatly enhance PML expression in vivo ( 9, 10). More recently, PML expression was found in many normal tissues, as well as in tumors of multiple origin, tumor progression being associated with loss of PML expression ( 11).
Exons 1 to 3, common to all isoforms, are translated into the tripartite motif called TRIM or RBCC (RING finger, B-box, coiled-coil; ref. 12). Within the RBCC core of PML, the function of the coiled-coil is clearly shown, e.g., multimerization ( 13). Exon 6 contains a nuclear localization signal that may be skipped, yielding cytoplasmic PML proteins ( 14). Consistent with the ability of RING (really interesting new gene) zinc fingers to bind E2 enzymes, several RBCC/TRIM proteins are E3 ubiquitin ligases ( 15– 18). The PML RING domain was shown to bind UBC9 ( 19), an E2 enzyme for the sumolation pathway. PML indeed harbors three sumolation sites: one in the RING finger, another in the first B-box (K160), and the last one in the nuclear localization signal (K490; refs. 20, 21). In the presence of arsenic trioxide, PML is sumolated on K160 and then degraded ( 21– 23).
The nuclear matrix–bound PML creates nuclear structures called PML nuclear bodies, nuclear domain 10, PML oncogenic domain, or Krüppel bodies whose functions are still unclear ( 24). These structures, which may recruit a wide variety of proteins, have been associated with many functions, including transcription, protein degradation, DNA repair, or telomere maintenance. The presence of many sumolated proteins on the PML bodies has also suggested a direct role in sumolation or in the accumulation of sumolated proteins. PML sumolation is not implicated in its nuclear matrix targeting, as initially proposed ( 23), but rather in the maturation of these domains, consisting of the recruitment of many partner proteins and acquisition of a specific shell-like morphology ( 21, 25– 27). The dynamic of PML localization between the nucleoplasm and nuclear bodies may reflect a sequestration function for PML, for which the most actively studied candidate has been the Daxx repressor ( 28– 32). Such a depot function for PML bodies could explain why alterations in their dynamics are found in a wide variety of pathologic situations such as viral infection, many types of stresses, or DNA damage ( 33).
Mice agenic for pml are viable, but have an increased rate of tumors, suggesting that pml could be a tumor suppressor ( 34). Ex vivo studies have also suggested that PML has proapoptotic and antiproliferative effects ( 24). Several studies have implicated TP53 in PML-triggered apoptosis or senescence ( 35, 36), although there is evidence that interference with the retinoblastoma, and not the TP53 pathway, allows escape from senescence of PML-IV-transduced cells ( 37, 38). Several mechanisms could link PML to TP53. CBP, an acetyltransferase, was proposed to acetylate TP53 in a PML-dependent manner ( 35). PML can also interact with a TP53 ubiquitin-ligase, Hdm2. After DNA damage, PML sequesters Hdm2 in the nucleolus resulting in the stabilization of TP53 ( 39, 40). Another mechanism linking PML to TP53 is the recruitment of Chk2 or HIPK2 kinases and TP53 in PML oncogenic domains under stress, which could facilitate the phosphorylation and activation of TP53 ( 41, 42). Recent studies have proposed a specific role for PML-III in the maintenance of centrosome stability ( 43). PML-III is targeted to the centrosome through an interaction with the kinase Aurora A. This interaction regulates the kinase activity of Aurora A, altering the checkpoint responsible for centrosome duplication. PML-III deficiency results in centrosome instability, possibly indirectly accounting for the genomic instability associated with PML loss ( 44) and the cancer susceptibility of pml−/− mice ( 45).
Alternative splicing of the COOH-terminal part of PML may be important, because most other RBCC/TRIM proteins display well-characterized domains in their COOH terminus (Supplementary Table S1). The multiplicity of the proposed roles for PML could reflect specialized functions of the distinct COOH-termini of the PML isoforms. To date, several motifs have been identified in the COOH terminus of PML: a nuclear export signal for PML-I ( 46) and destabilization motifs in PML-III and PML-IV ( 21, 47). PML-IV also harbors binding motifs for TP53, HDM2, and HDAC ( 32, 40). That the COOH-terminal sequence of PML-I is much more conserved from mouse to man than the RBCC motif could also be suggestive of an associated function, conveyed through a specific domain ( 48).
Here, we show that the five PML isoforms are all expressed in several human cell types, PML-I/II having the highest expression levels. Distinct subnuclear localizations were found when PML isoforms were expressed alone, suggesting that they make distinct interactions with specific cellular components. PML-I is a highly expressed conserved isoform and is likely to be an important contributor to PML function.
Materials and Methods
Cell lines, transfection, and treatment. HeLa, COS, PML-III transduced pml−/− MEF, SaOS, MRC5, PML-III expressing Chinese hamster ovary cell lines, and human macrophages were cultured in DMEM (Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum, 50 units/mL penicillin, 50 mg/mL streptomycin, and 2 mmol/L glutamine. NB4 and primary APL cells were cultured in RPMI (Life Technologies) supplemented with 10% fetal bovine serum, 50 units/mL penicillin, 50 mg/mL streptomycin, and 2 mmol/L glutamine. Cells were transfected either with the Effectene kit (Qiagen, Hilden, Germany) or LipofectAMINE 2000 (Invitrogen, Carlsbad, CA). Cells were treated with IFNα/γ (Roche, Mannheim, Germany) for 24 hours and 1 μmol/L of As2O3 for 8 to 12 hours. NB4 were treated with 1 μmol/L of all-trans retinoic acid (Sigma, St. Louis, MO) for 2 days.
Immunofluorescence. Cells were cultured on 12-mm diameter coverslips. Fixation was done either with 4% paraformaldehyde for 20 minutes at 4°C or methanol for 5 minutes at −20°C. Permeabilization was done with 0.1% PBS/Triton X-100. PML was labeled either with 5E10, a murine antibody directed against a common region for all the isoforms ( 49), or by isoform-specific rabbit sera (see below). The centrosome was labeled either with an antibody directed against γ-tubulin (GTU-88; Sigma) or with the monoclonal CTR453 directed against AKAP450, a marker of the pericentrosomal matrix ( 50). Bound antibodies were labeled with FITC or Texas red–conjugated secondary antibodies (Molecular Probes, Leiden, the Netherlands) for 1 hour and mounted with VectaShield + 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA). All the observations were made by confocal microscopy either with MRC1024 (60× lens; Bio-Rad, Hercules, CA) or LSM510META (63× lens, NA1.4 oil; Zeiss, Thornwood, NY). Electron microscopy was done as previously described ( 51).
Production and purification of glutathione S-transferase COOH terminus isoform–specific proteins. The isoform-specific COOH-termini of PML-I to PML-V were amplified by PCR, respectively, with the primers: 5′-CCCGGATCCGTGCAGGCAGCTGTGC-3′ and 5′-CCCGTCGACTCAGCTCTGCTGGGAGG-3′; 5′-CCCGAATTCTCCTCCATGGCTTCC-3′ and 5′-CCCGTCGACTCAGAGGCCTGCTTGACG-3′; 5′-CCCGGATCCGTCTCTTCCAGCCCTCAG-3′ and 5′-CCCGTCGACTCAGCGGGCTGGTGGGG-3′; 5′-CCCGAATTCGGGTTCTCCTGGGGC-3′ and 5′-CCCGTCGACCTAAATTAGAAAGGGGTGG-3′; 5′-CCCGGATCCGTGAGTGGCCCAGAAG-3′ and 5′-CCCGTCGACTCAATGCCTCACTGGAA-3′. The amplified DNA fragments, digested with EcoRI or BamHI on one side, and SalI on the other side, were inserted in pGEX4T1 at the corresponding sites. Protein expression was done in bacterial strains as BL21, pGroESL, AD494, and SG12060 at an absorbance of 0.4 to 0.8, for 4 hours at 25°C to 30°C with 1 mmol/L of isopropylthiogalactoside. The glutathione S-transferase (GST) purification was done as indicated ( 52).
Immunization of rabbits. Recombinant GST-PML fusion proteins (200 μg) were mixed with Freund's adjuvant and injected thrice into rabbits at an interval of 3 weeks. Two months after the first immunization, blood was drawn and serum obtained after coagulation. Purification of the specific antibodies was done as follows ( 52). Recombinant GST fusion proteins were fixed on CNBr-activated Sepharose (Amersham, Piscataway, NJ), following the manufacturer's protocol. After successive denaturation of the obtained resin with 100 mmol/L of glycine (pH 2.5), 10 mmol/L of Tris (pH 8.8), 100 mmol/L of triethylamine (pH 11.5), and 10 mmol/L of Tris (pH 7.5), the immune serum was diluted 10 times in 10 mmol/L of Tris (pH 7.5) before incubation with the GST recombinant protein-fixed Sepharose for 1 to 2 hours at 4°C. The resin was washed by 20 bed volumes of 10 mmol/L Tris (pH 7.5), and then by 20 bed volumes of 500 mmol/L NaCl, and 10 mmol/L of Tris (pH 7.5). The antibodies were eluted by 10 bed volumes of 100 mmol/L glycine (pH 2.5) and neutralized by 1 bed volume of 1 mol/L of Tris (pH 8). The depletion of the antibodies directed against the carrier GST was done by passage of the previously obtained antibodies on a GST-fixed Sepharose five times. Two distinct purifications of anti-PML-II antibodies were used in the studies reported here.
Nucleoplasmic extract, immunoprecipitation, and Western blot. The soluble fraction of the cells was extracted using radioimmunoprecipitation assay medium ( 53). Immunoprecipitation was done in radioimmunoprecipitation assay medium with the corresponding antibody and protein A-Sepharose (Sigma). The samples were then separated by SDS-PAGE. The immunoblots were incubated with peroxidase antibodies (PARIS) or peroxidase-soluble protein A (Amersham) for immunoprecipitation, and was revealed with an enhanced chemiluminescence kit (Amersham).
Pathology. Stored paraffin blocks of surgical samples removed for diagnostic procedures were used in this study after the diagnosis had been fully established. Patients were informed of the study according to the institution's regulations. Pathological lesions were selected for this study when the same paraffin block contained both an invasive carcinoma and surrounding preneoplastic lesions. Sequential 5-μm-thick sections were obtained on a microtome with water flow (HM 350 Niagara, Microm, Francheville, France). The subsequent sister sections were used for H&E staining and the successive different antibodies. An indirect immunoperoxidase method was done on a Ventana Nexes automatic, with antibodies directed against the five PML isoforms and the human TP53 protein (clone D07; Dako, Glostrup, Denmark) as primary antibodies, at a dilution of 1:100, without antigen retrieval.
Quantitative reverse transcription-PCR. RNA from bladder tumors, breast tumors, colon tumor, lymphoma, and white blood cDNA were obtained through the Hospital Saint-Louis, Paris, France. Brain and liver RNA were obtained from commercial sources. SaOS, H358, and HeLa RNA were extracted using TriZol reagent (Invitrogen). cDNA were obtained using Thermoscript kit (Invitrogen) using random oligomers as primers. Quantitative PCR were done using a Light Cycler (Roche). The primers and fluorescent probes were directed against either a common region to all PML isoforms (exons 2-3, Hs00231241_ml; Applied Biosystems, Foster City, CA) or the specific region for PML-I mRNA (exon 8a-9 junction: forward primer 5′ACCTCTGGTTTTCTTTGACCTCAAG3′, reverse primer 5′GAACTTGCTTTCCCGGTTCAC3′, probe: 5′ACAATGAAACCCAGAA GATT3′).
Validation of isoform-specific sera. Affinity-purified immune sera directed against PML-I to PML-V isoform-specific polypeptides (Supplementary Fig. S1) were first tested by immunofluorescence and Western blot on COS cells transiently transfected with expression vectors for each isoform. Using these sera, the expression of PML-I to PML-V was detected by immunofluorescence in the transfected cells (data not shown). Similarly, all isoform-specific sera, except anti-PML-I, efficiently detected its cognate protein on Western blots ( Fig. 1A ; data not shown). As expected, no cross-reaction between unrelated PML isoforms were observed in either immunofluorescence or Western blots. Three different antibodies [hen pan-PML, PGM3 (NH2-terminal epitope; ref. 54), and 5E10 (central epitope; ref. 49)] yielded identical results, demonstrating the absence of conformational bias (data not shown). These experiments therefore validate these sera and strongly suggest that the anti-PML-I sera contains antibodies that recognize highly conformational epitopes. Nevertheless, the PML-I protein could be detected with a serum directed against exon 8a, which is shared by PML-I and PML-IV ( Fig. 1A) and detects these two proteins with a similar efficiency.
Endogenous PML isoforms have distinct relative abundance. For each isoform-specific serum, we then detected nuclear dots in untransfected HeLa cells, that strictly colocalized with the Pan-PML monoclonal 5E10 (Supplementary Fig. S2). This experiment therefore shows that all endogenous PML isoforms are individually expressed and establishes that they all colocalize on nuclear bodies. Identical results were obtained with other cells such as macrophages or SaOS (data not shown).
When endogenous PML proteins were probed with a pan-PML antibody on a Western blot, a number of different proteins were detected ( Figs. 1 and 2 ). Interestingly, high molecular weight isoforms (which comigrate with PML-I/II, Fig. 1; data not shown), were clearly much more abundant than the low molecular weight ones (which comigrate with PML-III, PML-IV, and PML-V). Several distinct pan-PML sera, raised in rabbit, chicken, or mice, gave similar results in several cell lines or primary cells (data not shown). Yet, the fact that PML may be sumolated and the existence of specific sites for proteases ( 55), might complicate isoform identification exclusively based on their size. Probing replicas of these Western blots with the different isoform-specific antisera, we could directly confirm the nature of the PML proteins that was inferred from their size ( Fig. 1B), hence, demonstrating the relative abundance of the higher molecular weight isoforms compared with the smaller ones. To directly define the relative abundance of PML-I and PML-IV, we used the anti-PML-I+IV sera in a number of cell extracts derived from primary human macrophages or MRC5 cells ( Fig. 1C and D). An abundant protein that comigrates with PML-I was observed, together with a 5- to 10-fold less abundant protein comigrating with PML-IV ( Fig. 1C and D). In conclusion, the longer isoforms, PML-II/PML-I, are the most abundant ones, followed by PML-III/IV/V, whereas PML-IV is far less abundant than PML-I. Consistent with the colocalization of all endogenous PML isoforms in PML bodies, when we immunoprecipitated endogenous PML-IV from SaOS cells, we could coprecipitate a higher molecular weight PML isoform (I or II), directly demonstrating that endogenous isoforms not only colocalize, but are also physically associated in vivo ( Fig. 1E).
Furthermore, to show the existence of the various PML isoforms in other tissues, we analyzed primary cultures (macrophages) or cell lines (HeLa, NB4, MRC5, HCT116, and ME180) with the sera obtained above. All isoform-specific sera detected proteins that were up-regulated by IFN and degraded upon arsenic exposure, a hallmark for PML ( Fig. 2A; data not shown; refs. 21, 22). Again, using the pan-PML sera, protein species comigrating with PML-I and PML-II were much more abundant than the more rapidly migrating species ( Fig. 2A; data not shown). Therefore, the transformed or the differentiation stage does not seem to affect the ratio of isoform expression. Although in HeLa and NB4 cells, all isoforms were very efficiently degraded upon arsenic treatment; intriguingly, PML degradation was not observed in human macrophages or in differentiated NB4 cells. Rather, in those cells, we observed a set of high molecular weight proteins, likely corresponding to poly-modified PML/SUMO proteins ( 22, 23). Arsenic first triggers PML targeting to the nuclear matrix, then its modification by SUMO1/2/3, and finally, its proteasome-dependent degradation ( 21). This last step was not observed in slowly dividing cells, such as macrophages or differentiated NB4 cells, suggesting that progression through the cell cycle is required for the efficient degradation of sumolated PML.
PML-I is encoded by abundant mRNA in primary cells. The comigration of PML-I and PML-II precludes the determination of the most abundant isoform by Western blotting. PML-I abundance was indirectly evaluated through the comparison with PML-IV on Western blot (see above). We decided to determine the relative abundance of PML-I mRNA in various tissues, tumors, and cell lines. As shown in Fig. 2B, PML-I mRNA represents a very significant fraction (40-80%) of the total PML mRNAs in nontransformed cells (MRC5 and WI-38) or normal tissues. In contrast, PML-I mRNA has a low abundance in transformed cell lines (SaOS, H358, and HeLa). These observations are consistent with the results from the Western blot and further suggest that loss of PML-I expression may correlate with transformation, as suggested for pan-PML in vivo ( 11).
PML-III is not associated with the centrosome. In an attempt to confirm the colocalization of PML-III and the centrosome, we explored various cellular situations associated with different levels of PML-III expression ( Fig. 3 ). PML-III (detected with either PML-III or pan-PML antibodies), stably expressed in Chinese hamster ovary or pml−/− MEF showed no detectable colocalization with the centrosome. Similarly, HeLa cells, MRC5 human primary fibroblasts, APL-derived cell line NB4, or APL primary cells never showed any detectable centrosomal labeling with several pan-PML or anti-PML-III antibodies, suggesting that the centrosome does not significantly accumulate PML-III.
Expression of PML isoforms in tumors. Because PML has been implicated in cell transformation and senescence, and has been shown to be overexpressed in many tumors, and then lost with tumor progression ( 9– 11), we examined the expression of PML isoforms in human cancers. We first validated isoform-specific antibodies on skin biopsies from patients with psoriasis. This condition is associated with a local secretion of IFNα, which induces high PML expression ( 5, 9). A very high expression of all PML isoforms was easily detected as multiple nuclear dots for all six sera (PanPML, PML-I to PML-V; data not shown). Hence, all isoforms are expressed in vivo and our sera (including PML-I) could efficiently detect them. A panel of five breast cancers and five skin carcinomas were then analyzed with all isoform-specific PML antibodies, as well as TP53 (data not shown). Interestingly, only the three tumors in which TP53 was detectable expressed PML (one breast tumor and two skin cancers), consistent with the transcriptional induction of PML by TP53 in vivo. Among these three tumors, all PML isoforms were similarly expressed in patches of epithelial cancer cells, but not in the stroma. Interestingly, in one of the samples, a strong cytoplasmic staining, together with rare nuclear bodies, was found with all isoforms ( Fig. 4 ). PML was recently proposed to have a cytoplasmic function and our demonstration of cytoplasmic PML in vivo could lend support to this hypothesis. We cannot rule out that the cytoplasmic staining results from tissue fixation, although TP53 was strictly nuclear in the subsequent sister slides, and preimmune sera failed to label any of these cells (data not shown).
Stably expressed PML isoforms each accumulate in distinct nuclear domains. We then expressed each PML isoform in primary or immortalized pml−/− embryonic fibroblasts and analyzed PML localization by immunofluorescence. Strikingly, distinct patterns of localization were found ( Fig. 5 ). Whereas PML-I has both a cytoplasmic and a nuclear distribution (consistent with the presence of a nuclear exclusion signal), PML-II had a thread-like distribution. In contrast, PML-IV expression yielded a large number (>20) of small irregular bodies scattered through the nucleus. Finally, PML-V yielded very large and dense bodies ( Fig. 5). These observations suggest that each isoform interacts with different nuclear components, which contribute to these specific subnuclear localizations. Note, however, that when several isoforms are coexpressed, they always colocalize, most likely due to their ability to heterodimerize ( Fig. 1E; ref. 13).
We similarly stably expressed the different PML isoforms in Chinese hamster ovary cells and analyzed their localization by immunoelectron microscopy. Again, different localizations were consistently found for each isoform ( Fig. 6 ). The definition of the electron microscopy highlights localizations that were not apparent from the immunofluorescence. Many of these localizations are distinct from the well-described PML-III-associated structures ( 49). In particular, PML-IV was occasionally found in the dense fibrillar compartment of the nucleolus ( Fig. 6F), PML-II generated very thin bubble-like structures at the immediate vicinity of the nuclear membrane, as well as a fine inner lining of the nuclear membrane ( Fig. 6D and E), whereas PML-V yielded nuclear bodies with a characteristic large shell ( Fig. 6G). Finally, in several cells, some cytoplasmic PML-I was found adjacent to mitochondria ( Fig. 6C; data not shown).
Not all PML isoforms are conserved between human and mouse. The genomic organization of the murine pml gene was previously published ( 48), revealing a striking similarity in the COOH terminus of human and mouse PML-I. Using a set of primers on the different identified exons, together with oligo-dT primers, and PCR-amplifying pml transcripts, we could sequence new murine pml cDNAs and align them on a genomic clone that comprises exons 6 to 9 of the murine pml gene (NT039479.1/Mm39514.30). We found an equivalent of the 5′ part of human exon 7ab that encodes a protein highly similar to PML-V (although the sequenced contig contains an in-frame stop codon). Yet, we could never detect an equivalent of human 7b or 8b, suggesting that the mouse has no orthologue of PML-II or PML-III (Supplementary Fig. S3). Although the isoform-specific sequence of PML-IV is only 12 amino acids long, three large hydrophobic residues were conserved, possibly reflecting some sequence conservation in this segment.
Here we show the expression of all five PML protein isoforms previously identified by cDNA cloning and confirm that they have different relative levels of expression. We have not been able to find a situation in which significant differences in the relative abundance of PML isoforms exist and, at large, similar patterns of protein isoform expression were found in transformed or primary cells. A number of studies have shown altered levels of PML expression in tumors ( 9– 11). The major determinants of the transcriptional regulation of PML are type I and II IFNs ( 5, 56), as well as TP53 activation ( 6). It is therefore not surprising that PML is overexpressed in tissues undergoing inflammation, stress, or oncogenic transformation. Loss of PML expression upon tumor progression ( 9, 11) may reflect impairment of TP53 activity. We analyzed the expression of PML-I mRNA in cells of various origins and found that its expression was highest in normal cells, and was lost in highly transformed cell lines ( Fig. 2A). Analysis of other isoforms by quantitative reverse transcription-PCR was not possible because the specific exon junctions for many of them are shared with mRNA encoding other isoforms.
At the cell biology level, expression of the distinct PML isoforms in a variety of cellular backgrounds consistently yielded distinct localizations, as recently shown in another setting ( 57). We have not been able to confirm the specific localization of PML-III to the centrosome ( 43). Rather, we found that PML-III colocalizes with other isoforms on typical nuclear bodies, consistent with our previous observations. Similarly, that we have not been able to find a mouse homologue to the isoform-specific region of PML-III does not argue for a critical role of PML-III (Supplementary Fig. S3). PML-I has both a nuclear and a cytoplasmic distribution, consistent with the presence of a nuclear exclusion signal. Previous studies have indeed shown that in a variety of settings (APL, virus-infected cells, expression of specific splice variants, etc…), PML may have a cytoplasmic localization ( Figs. 5 and 6; Supplementary Fig. S4; refs. 14, 49, 58). In several tumor samples, we found a cytoplasmic expression of PML, which may reflect an enhanced cytoplasmic sequestration of PML or a greater relative expression of PML-I ( Fig. 4; data not shown). The thread-like images found with PML-II were previously observed in cells transformed with RasV12 ( 59). Interestingly, the threads observed on adenovirus infection were shown to be dependent on the expression of PML-II ( 60– 62). The thin bubbles on Fig. 6D could correspond to sections of these threads. The lamin-like images observed in immunoelectron microscopy could suggest some association of PML with the nuclear membrane, as previously proposed for cytoplasmic membranes of early endosomes ( 14). Altogether, these distinct localizations strongly suggest that the specific COOH-terminal ends of these isoforms are involved in interactions with cellular components, that will impose some constraints on PML localization.
PML-IV, the most intensively studied isoform to date, is expressed at low levels, at least when compared with PML-I and PML-II. Enforced expression of PML-IV will yield homodimers, which are unlikely to occur with endogenous PML proteins. Interestingly, overexpression of human PML-IV in pml+/+ MEF, but not in pml−/− MEF, triggered senescence ( 63), which could suggest that only heteromultimers are active to promote senescence. Together with the fact that PML-I contains all conserved coding exons, its relative abundance could suggest that PML-I is the prototype and the others are splice variants. Paradoxically, this specific isoform has virtually not been studied to date and it would be interesting to compare its properties with those of PML-IV, in particular, with respect to the direct or indirect control of TP53 function.
Grant support: Eli Lilly Foundation and Ligue Contre le Cancer. Wilfried Condemine and Yuki Takahashi had a scholarship from Association pour la Recherche sur le Cancer and bourse du Ministère de la Recherche et des Technologies, as well as the Virus Cancer Prevention Association.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. Zina Doubeikovski for the protocol of GST protein purification and the elaboration of anti-PML-I/-IV serum; Anne-Marie Poorters for the sequencing of constructs, Bernard Boursin for the elaboration of the figures of the article; Dr. Marie-Claude Guillemin and Hassane Soilihi for the immunization of rabbits and for collecting their blood; and Frédéric Brau, Michel Schmidt, Micaël Yagello, and Niclas Setterblad at the Services Communs d'Imagerie of the IUH of Saint-Louis Hospital for the confocal microscopy.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
↵4 Y. Takahashi, unpublished data.
- Received October 20, 2005.
- Revision received April 7, 2006.
- Accepted April 19, 2006.
- ©2006 American Association for Cancer Research.