PML bodies: a meeting place for genomic loci?

The mammalian nucleus is a complex organelle organized into chromatin territories and discrete nuclear compartments or bodies. One of these is the promyelocytic leukemia (PML) body, also known as the PML oncogenic domain (POD), nuclear domain 10 (ND 10) or Kremer (Kr) body. There are approximately 5-30 bodies observed per nucleus, ranging in size from ∼0.2 to 1 μm (Melnick and Licht, 1999). The importance of PML bodies in cell differentiation and cell growth was first indicated in studies of promyelocytes from patients suffering from acute promyelocytic leukemia (APL), in which a fusion of the PML protein and the retinoic acid receptor α (RARα) occurs as a result of the chromosomal translocation t(15:17) (de Thé et al., 1991; Goddard et al., 1991; Kakizuka et al., 1991). In promyelocytes from these patients, the PML bodies are disrupted. When cells are treated with all-trans retinoic acid (ATRA) or with arsenic trioxide, PML bodies reform, and the APL phenotype is reversed in patients treated with these agents (reviewed by Melnick and Licht, 1999).

When the importance of PML bodies in cell differentiation and cell growth was first determined, the components of this nuclear structure were still unknown. Later, several groups demonstrated that the PML protein is a constituent of PML bodies (Dyck et al., 1994; Koken et al., 1994; Weis et al., 1994). The identification of other proteins associated with PML bodies has provided clues to the function of this subnuclear compartment (Table 1) (reviewed by Negorev and Maul, 2001; Eskiw and Bazett-Jones, 2002). However, over 60 proteins are known to localize to this compartment (see the Nuclear Protein Database http://npd.hgu.mrc.ac.uk/) (Dellaire et al., 2003), implicating these structures in virtually every nuclear activity, including transcription (reviewed by Zhong et al., 2000), DNA repair (reviewed by Dellaire and Bazett-Jones, 2004), apoptosis (reviewed by Takahashi et al., 2004), tumor suppression (reviewed by Salomoni and Pandolfi, 2002), proteolysis (Lallemand-Breitenbach et al., 2001), and the antiviral response (reviewed by Regad and Chelbi-Alix, 2001).

Despite the attention paid to these structures, the function of PML bodies is still not fully known, and three models have been proposed. In the first model, the bodies are proposed to be aggregations of excess nucleoplasmic protein (reviewed by Negorev and Maul, 2001). The concentration of a nuclear protein through formation of aggregates would provide a mechanism for regulating nucleoplasmic levels of that protein, in which PML bodies would release proteins as they are needed. In the second model, PML bodies are proposed to be sites of post-translational modification and degradation of proteins. Observations supporting this model include the acetylation and phosphorylation of p53 at PML bodies, which enhances the activity of p53 (Pearson et al., 2000; D'Orazi et al., 2002; Hofmann et al., 2002), and the localization of the 19S and 20S proteasome subunits at some PML bodies (Lafarga et al., 2002). In the third model, PML bodies are proposed to be sites of specific nuclear activities, such as transcriptional regulation and DNA replication. Evidence for this model includes the detection of nascent RNA around PML bodies (Boisvert et al., 2000), the association of PML bodies with regions of high transcriptional activity (Wang et al., 2004), and the non-random nature of PML body assembly (based on the conservation of their size and position) following dissociation and re-formation as a result of cellular stress (Eskiw et al., 2003). These three models of PML body function need not be mutually exclusive.

Here, we discuss the possibility that PML bodies interact with specific genes or genomic loci. Examination of the assembly and disassembly of PML bodies after cell stress, the role of PML bodies in the transcription and replication of DNA viral genomes, and the association of PML bodies with regions of high transcriptional activity in the human genome, all support the hypothesis that the contacts seen between PML bodies and the surrounding chromatin in situ are in fact functional.

A central question is whether PML bodies themselves are sites of specific nuclear activity that PML proteins and other PML-body-associated proteins do not perform elsewhere in the nucleoplasm. To date, only indirect evidence supports the idea that the bodies are sites of activity. PML bodies can remain in the same position in the nucleus for long periods of time (Wiesmeijer et al., 2002; Eskiw et al., 2003) unlike other large accumulations of nuclear protein, such as matrix-associated deacetylase (MAD) bodies (Downes et al., 2000). MAD bodies are spherical aggregates formed by over-expression of histone deacetylases; they can be similar in size to PML bodies (100 nm-1 μm diameter) but are highly mobile in the nucleoplasm, moving at speeds as high as 1 μm/second (Kruhlak et al., 2000). We hypothesize that PML bodies are positionally stable because they interact with structures in the nucleoplasm (rather than because of their size) (Eskiw and Bazett-Jones, 2002). However, PML bodies do exhibit a periodic, energy-dependent, oscillatory movement of up to one body diameter in amplitude (Eskiw et al., 2003). Thus, movement of PML bodies might be restricted by functional contacts with other nuclear substructures, such as chromatin.

Positional stability of PML bodies could also depend on contacts with non-chromatin structures. PML bodies are found in the nuclear matrix, which is the insoluble fraction of the nucleus that remains following digestion of nucleic acid and extraction with high concentrations of salt and detergent (Pederson, 1998). Although there is no ultra-structural evidence for a nuclear skeleton in intact nuclei, a structure analogous to the cytoskeleton might be present. Dynamic studies of various nuclear components, such as chromatin (Chubb et al., 2002) and supramolecular complexes including mRNPs (Shav-Tal et al., 2004), PML bodies (Eskiw et al., 2003), MAD bodies (Kruhlak et al., 2000) and Cajal bodies (Platani et al., 2002), are inconsistent with a fibrous network. However, short fibers of polymerizing and depolymerizing actin or lamin, rather than a static network of these filaments, might yet be shown to help provide a localized architecture that facilitates the organization of functional domains within the nucleus.

Energy-filtered transmission electron microscopy (EFTEM), also known as electron spectroscopic imaging (ESI) (Bazett-Jones and Hendzel, 1999) (Fig. 1), reveals PML bodies in mid- or late-G1 phase or in G2 phase to be completely surrounded by chromatin and/or accumulations of ribonucleoprotein (RNP) fibers and granules. The chromatin surrounding the bodies exists as both extended 10 nm and 30 nm fibers, but higher-density chromatin can also be found as irregular arrays of tightly packed 30 nm fibers. Since the chromatin itself is relatively immobile, we propose that the mobility of PML bodies is restricted by contacts with the surrounding chromatin domains. In addition, direct contacts link the surrounding chromatin and the surface of the PML body, some of which are protein-based fibers extending from the core towards the chromatin (Eskiw et al., 2003) (Fig. 1).

The integrity and stability of PML bodies are lost when chromatin is disrupted by stress, transcriptional repression or early apoptotic events. Under such conditions, PML bodies become mobile (Maul et al., 1995; Eskiw et al., 2003; Nefkens et al., 2003; Eskiw et al., 2004). The immediate increase in mobility within minutes of the addition of an exogenous nuclease strongly argues that a chromatin-dependent mechanism underlies PML body stability, because protein-based contacts would not initially be disrupted by nuclease treatment (Eskiw et al., 2004). The redistribution of PML bodies into numerous, smaller, PML-protein-containing microstructures or microbodies occurs through fission events (Eskiw et al., 2003; Eskiw et al., 2004). Microstructures by definition contain the PML protein but lack other PML-body-associated proteins, such as Sp100 or SUMO-1, whereas microbodies are biochemically indistinguishable from the parental PML bodies.

The fission of PML microstructures or microbodies occurs as chromatin retracts away from the bodies, which might pull supramolecular PML protein accumulations away from the parental body. Alternatively, or in addition, the loss of chromatin from the periphery of the body might shift the dynamic equilibrium of supramolecular PML protein bound to the body, favoring the loss of PML protein from the body by mass action.

Following heat stress, >50% of PML protein immediately dissociates from parental PML bodies by fission events, but no new PML bodies form from this release. The disassembly and re-formation of PML bodies after heat stress might be a result of the dynamic nature of the constituent proteins. It has been shown that CBP, PML protein and Sp100 (components of PML bodies) are highly dynamic and exchange rapidly between the nuclear body and the surrounding nucleoplasm (Boisvert et al., 2001; Wiesmeijer et al., 2002). In fact, after recovery from the stress (2-4 hours), the PML bodies reform in the original locations and regain their original relative sizes. This argues that the assembly of PML bodies is non-random. The conservation of size following disassembly and re-formation could also reflect the size of the space within the chromatin-based domain in which the body is found, as well as the number of chromatin contacts that are available for interaction with PML protein or PML-associated factors. Thus, chromatin might affect the size and positional stability of nuclear compartments, such as the PML body, and the dynamic nature of the proteins constituting PML bodies might affect the disruption and re-formation of this nuclear compartment.

Contacts between chromatin and PML bodies could have a role more profound than just the maintenance of their positional stability. One hypothesis is that the disruption of PML bodies might directly affect the availability of PML-body-associated proteins within the nucleus and release these proteins as PML bodies are disrupted. It is equally possible that the association of PML bodies with chromatin also brings together specific gene loci regulated by PML-body-associated proteins. The genes brought into the vicinity of the PML body might then be jointly regulated through clustered co-activators such as CBP [a co-activator and a histone acetyltransferase (HAT)], and co-repressors, such as Sp100 and the histone deacetylases (HDACs) (Bloch et al., 1996; Bloch et al., 1999; Bloch et al., 2000; Boisvert et al., 2000). Such a role for PML bodies is reminiscent of that proposed for matrix attachment regions (MARs) in DNA (Cockerill and Garrard, 1986). Future studies to determine whether the chromatin that contacts PML bodies actually contains MARs might help to determine whether PML bodies, at least in part, provide the `matrix' to which these DNA sequences attach.

Below we discuss the localization of viral and genomic loci to PML bodies. All the studies referred to used fluorescence in situ hybridization (FISH) techniques. This technique allows detection of specific loci by fluorescence microscopy. Its weakness is the limited spatial resolution provided by the light microscope. The question that must be kept in mind, therefore, is whether the presence of specific loci within a few diameters of a PML body really reflects a physical association, which implies a functional relationship.

PML bodies might regulate the transcription and replication of foreign DNA introduced into the nucleus during viral infection. Early work showed that herpes simplex virus type 1 (HSV-1) forms replication sites that display a non-random distribution in infected binucleated cells (de Bruyn Kops and Knipe, 1994), implying that the virus either creates replication sites de novo or uses a pre-existing nuclear compartment to assemble its replication complexes. Maul et al. later demonstrated that PML bodies are the nuclear compartments responsible for the non-random distribution of HSV-1 replication sites by showing that the parental HSV-1 DNA localizes to positions adjacent to PML bodies and that viral DNA replication is closely associated with PML bodies (Maul et al., 1996). This observation is not unique to HSV-1: adenovirus 5 (Ad5), simian virus 40 (SV40) and human cytomegalovirus (HCMV) also deposit their genomes adjacent to PML bodies (Ishov and Maul, 1996). Moreover, in HCMV infection, only the genomes deposited adjacent to PML bodies are transcribed (Ishov et al., 1997).

These past studies show that viral genomes are targeted to PML bodies. Since PML bodies are positionally stable for long periods of time (Eskiw et al., 2004), it might be concluded that viral genomes migrate to PML bodies. A recent study has shown that, when cells are infected with an HSV-1 mutant that does not disrupt PML bodies, PML-body-like complexes appear to form adjacent to incoming viral nucleoprotein complexes (Everett et al., 2004). Furthermore, the researchers showed that only in productive infections are PML bodies associated with the viral genome. A possible hypothesis to explain the association of viral genomes with PML bodies is that the high transcriptional activity of the viral genomes causes the deposition of PML body proteins at the active sites.

Since PML bodies define the location of early transcription and replication of at least four DNA viruses, what are the components needed for this localization? In the case of SV40, the minimum requirement is the SV40 core origin of replication and the T-antigen protein (T-ag) (Tang et al., 2000). Similarly, the minimum requirement for localization of the HSV-1 genome to PML bodies is the viral origin of replication OriS, the viral immediate-early proteins ICP4 (for `infected cell protein') and ICP27, and Daxx (Tang et al., 2003). The localization of HCMV genomes and transcripts to sites adjacent to PML bodies has been reported to be dependent on the trans-activator tegument protein pp71 (Ishov et al., 2002). However, Ishov et al. did not perform FISH analysis studies using the HCMV genome but showed indirectly that HCMV genomes associate with PML bodies by relying on IE2 as a surrogate marker for these genomes. Although IE2 is a factor that defines the immediate-early transcript environment of HCMV (Ishov et al., 1997), caution must be taken in interpreting these results until FISH analysis can confirm the true localization of the HCMV genome in the presence or absence of pp71.

The common theme among these three examples is that the proteins involved in targeting to PML bodies are also required for viral transcription or replication. Therefore, PML bodies might be preferred sites for the transcription and replication of DNA viruses, which would support the hypothesis that these bodies are sites of gene activity. Below, we examine the evidence supporting a role for PML bodies in transcription and replication of DNA viral genomes.

If the localization of DNA virus genomes to PML bodies does facilitate the transcription of viral genes, by what mechanism is transcription facilitated? Ishov et al. have proposed the `immediate transcript environment model' (Fig. 2) (Ishov et al., 1997). In this model, RNA transcribed from HCMV genomes deposited adjacent to PML bodies accumulates in adjacent SC35 domains, which are the nuclear compartments in which splicing factors accumulate and/or nascent transcripts are processed (Huang and Spector, 1991). Once the immediate-early viral protein IE72 is translated, it too is targeted to PML bodies and leads to their subsequent disruption. It was shown that IE86 localizes to the junction between PML bodies and SC35 domains, and recruits transcription factors such as TBP and TFIIB to this junction. Both IE72 and IE86 are transcriptional activators of early cytomegalovirus gene expression (Malone et al., 1990; Stenberg et al., 1990). HSV-1 produces a similar spatial organization to PML bodies, SC35 domains and viral transcripts (Ishov et al., 1997). The deposition of viral transcripts thus seems to be a result of transcription at PML bodies, which supports the hypothesis that PML bodies are sites of gene activity, in this case transcription.

PML bodies might also participate in viral DNA replication. The replication of the HSV-1 genome has been examined in some detail. From initial work by Weller and colleagues, it was assumed that, during early HSV-1 infection, PML bodies are disrupted and that PML protein accumulates in viral replication compartments as infection progresses (Burkham et al., 1998; Burkham et al., 2001). This model of HSV-1 replication compartment development requires PML body disruption for the progression of viral infection. However, several studies do not support this conclusion. For example, HSV-1 infection can progress without PML body disruption when an ICP0-null HSV-1 strain is used (Maul et al., 1996). Second, using the same ICP0-null strain of HSV-1, Everett and co-workers showed that PML bodies reorganize as replication compartments develop, but no PML protein is found in these compartments (Everett et al., 2004; Everett and Zafiropoulos, 2004). In addition, PML bodies remain intact during productive infection with another double-stranded DNA virus, SV40 (Ishov and Maul, 1996). Thus, it appears that the disruption of PML bodies might facilitate, but is not absolutely required for, replication of double-stranded viral DNA.

The disruption of PML bodies in HSV-1 infection could prevent the establishment of the antiviral state (Chee et al., 2003), which is an interferon-driven process in which PML protein impairs viral replication (Regad and Chelbi-Alix, 2001; Chee et al., 2003). However, the actual mechanism of PML body disruption differs between virus types. For example, in HSV-1 infection, ICP0 disrupts PML bodies (Maul and Evertt, 1994). ICP0 possesses E3 ligase activity (Van Sant et al., 2001; Boutell et al., 2002) and causes the proteasome-dependent degradation of PML protein (Everett et al., 1998). In HCMV infection, IE72 disrupts PML bodies by dispersing PML protein from the nuclear compartment (Korioth et al., 1996). The E4-ORF3 protein of Ad5 disrupts PML bodies by redistributing the PML protein into tracks (Doucas et al., 1996; Ishov and Maul, 1996). By contrast, PML bodies in cells infected with SV40 are not disrupted (Ishov and Maul, 1996).

Recent live-cell studies support the idea that PML bodies might be required for HSV-1 replication. For example, replication compartments form only from HSV-1 genomes associated with PML bodies (Sourvinos and Everett, 2002). These replication compartments develop from ICP4-containing foci (Everett et al., 2003) and, at the same time, new PML-containing structures form adjacent to these foci (Everett et al., 2004). Although it was initially concluded that HSV-1 viral genomes replicate more efficiently in association with PML bodies (Sourvinos and Everett, 2002), it appears that the de novo formation of body-like structures containing PML protein preferentially occurs at sites of efficient viral replication (Everett et al., 2004). Therefore, the disruption of PML bodies by proteins of some double-stranded DNA viruses might enhance productive infection, perhaps by repressing a PML-body-dependent antiviral response, whereas the redistribution of PML protein into new PML-protein-containing structures appears to enhance viral replication by facilitating the establishment of replication compartments.

Although instructive as a paradigm for the role of PML bodies in gene replication and transcription, viral infection is clearly not a normal state of cellular homeostasis. Thus, observations with respect to the association or nucleation of PML bodies at sites of viral genome expression and replication might not necessarily reflect the dynamics or functional relationship between PML bodies and cellular chromatin under normal growth conditions.

Given that PML bodies seem to play a role in viral gene expression and DNA replication, they might also function in the transcriptional regulation of cellular genes. Many examples of the non-random association of PML nuclear bodies in mammalian cells with specific chromosomal loci or chromatin have been reported. For example, Grande et al. found that PML bodies are adjacent to domains of replicating DNA in middle-late S-phase of the cell cycle (Grande et al., 1996). Providing another example, Tsukamoto et al. showed that multiple copies of a plasmid containing the binding sites for LacI and TetR, when integrated into the genome, associate with PML bodies regardless of transcriptional status (Tsukamoto et al., 2000). However, the integration of large repetitive arrays of prokaryotic DNA such as the Tet and Lac operator sequences, and overexpression of the DNA-binding proteins required for visualization of chromatin containing these arrays, might not provide a faithful approximation of normal chromatin behavior or associations within the mammalian nucleus.

To date, the combined immunolocalization of PML bodies with the localization of specific gene loci by immuno-FISH has provided the most convincing data to support the non-random association of PML bodies with specific chromosomal loci. For example, immuno-FISH analysis of Jurkat T cells identified TP53 sequence on the surface of a PML body in ∼50% of cells analyzed, whereas the BCL2 locus was never associated with PML bodies (Sun et al., 2003). TP53 encodes the p53 tumor suppressor, which promotes cell-cycle arrest or apoptosis in response to cell stress and accomplishes this in part by regulating the expression of genes such as p21, Mdm2, Bax and PML (Prives, 1998; Fridman and Lowe, 2003; Willis et al., 2003; de Stanchina et al., 2004). Data from several sources indicate that PML bodies play a central role in potentiating p53-mediated pathways, possibly by assembling complexes that positively and negatively regulate p53 function (Bernardi and Pandolfi, 2003; Dellaire and Bazett-Jones, 2004). It is also possible that PML bodies regulate the concentration of factors within the local chromatin environment that influence TP53 gene expression. Because p53 accumulates and is activated at PML bodies (Pearson et al., 2000; Fogal et al., 2000), it is tempting to speculate that, in addition to the TP53 locus, this sub-nuclear compartment also regulates p53-responsive promoters within the vicinity of the body.

Immuno-FISH analysis also indicates that the major histocompatibility (MHC) class I gene cluster on chromosome 6 in human primary fibroblasts is associated with PML bodies (Shiels et al., 2001). Statistical analysis of the mean minimal distance (MMD) between PML bodies and the MHC class I genomic loci indicates a non-random association with PML bodies. This region contains several genes that are constitutively expressed in fibroblasts, including LMP2, LMP7, Tap1 and Tap2. The MHC class I gene cluster lies at the centromeric end of chromosome 6 and extends over 1.6 Mb. By contrast, a gene-poor region of chromosome 6 [such as the 500 kb region of chromosome 6 (6p24)] has a significantly lower degree of association with PML bodies. Insertion of a fragment of the MHC locus into chromosome 18 increases the association of this chromosome with PML bodies (Shiels et al., 2001). However, other gene-rich regions, such as the epidermal differentiation complex (EDC) located on chromosome 1, do not significantly associate with PML bodies (Shiels et al., 2001). Although the EDC region is very gene rich (27 genes in 2.5 Mb of DNA), only five of these genes are known to be expressed in fibroblasts. Therefore, it is unclear whether gene-rich regions alone or the transcriptional activity of particular genes within that domain contribute to the association of PML bodies with these loci.

Wang et al. recently examined the importance of transcriptional activity versus gene density for the association of 54 different genomic loci from 10 chromosomes (1, 6, 7, 9, 14, 16, 17, 18, 19 and X) with PML bodies in human fibroblasts by immuno-FISH (Wang et al., 2004). In this study, they compared the MMD of each locus with that of the LMP/TAP region of chromosome 6 (considered to associate with PML bodies significantly) (Shiels et al., 2001). Gene-rich chromosomal regions showed significant association with PML bodies compared with gene-poor regions. Although PML bodies appeared to associate with regions of high transcriptional activity, the actual level of transcription of an associated locus was less important for this association. For example, genes such as Notch 1, which have low transcriptional activity in fibroblasts but reside in regions of chromatin that have high transcriptional activity, can associate with PML bodies in these cells.

PML bodies also appear to be more associated with genes of the active X chromosome than those of the inactive X chromosome in female cells (Wang et al., 2004). For example, 6 of 11 genes on the active X chromosome (G6PD, GLA, LAMP2, UBE2A, NAP1L3 and PSMD10) exhibit a more significant association with PML bodies than their counterparts on the inactive X chromosome. By contrast, the centromeres and several genes in the pseudo-autosomal region of both X chromosomes (ALTE, DMD and Usp9X on the Xp arm) do not show significant differences in their association with PML bodies. Similarly, the XIST loci of both X chromosomes associate with PML bodies to the same extent.

Finally, the association of specific loci with PML bodies might also be affected by the cell cycle. The core histone gene cluster on chromosome 6 is more closely associated with PML bodies in S phase, when histones are transcribed, than in G₁/G₂ phase cells, when these genes are silent (Shiels et al., 2001).

PML bodies thus seem to associate with chromosomal regions of high transcriptional activity and/or gene density, and in some cases the association may be cell-cycle dependent. Yet, not all transcriptionally active regions of the genome associate with PML bodies, which suggests that the specificity of PML body association is determined by locus-specific factors rather than by the transcription level. The identities of these specific factors and the PML body components with which they interact remain to be elucidated. The studies discussed above also indicate that PML body position might be non-random, and we suggest that PML bodies act as anchoring points for specific chromatin regions and perhaps have a function analogous to that postulated for MARs. More experiments are required to address this.

PML bodies are positionally stable over long periods during interphase. Following stress, they are conserved in size and position after a cycle of disruption and re-formation. We conclude that their position and size is non-random and might be dictated by chromatin, which might restrict their mobility. For example, chromatin and protein contacts can be observed by electron microscopy between the surrounding chromatin in the vicinity of the body and the surface of the body (Fig. 1). PML bodies are also sites of viral gene transcription and replication, as well as being associated with particular genomic loci, including TP53 and the MHC I and histone gene clusters. These observations support a role for PML bodies in coordinating gene activity (Fig. 3).

How PML bodies come to associate with particular gene loci is not clear, and at least two mechanisms for their formation at these sites have been evoked. In the first mechanism, PML-body proteins accumulate at sites of high transcriptional activity, as seen during viral infection. However, as discussed above, viral infection is not a normal cellular state. In the second mechanism, genomic loci are recruited to PML bodies and their transcriptional activity is regulated, which would be reminiscent of what is seen during the recruitment and silencing of genomic loci at the heterochromatin compartment containing the Brown locus in Drosophila (Csink and Henikoff, 1996; Dernburg et al., 1996) and the Ikaros complex in maturing B lymphocytes (Brown et al., 1997). These two mechanisms need not be mutually exclusive. However, the nucleation of PML bodies at specific genomic loci in mammalian cells has yet to be demonstrated.

The definition of putative associations between cellular genomic loci and PML bodies has relied heavily on the criteria used to determine non-random association. To date, most of the data supporting the association of specific loci with PML bodies rely on the interpretation of observations made by immunofluorescence microscopy. The lack of sufficient spatial resolution of the fluorescence microscope makes it difficult to define a non-random association. The application of chromatin immunoprecipitation techniques to support immuno-FISH analysis when defining functional interactions between PML body components and specific genes could help resolve this issue. Furthermore, one might expect that, if PML bodies are involved in organizing transcriptional domains, they might associate with genomic loci in late mitosis or early G1 phase of the cell cycle, when the organization of chromosomal domains is re-established. Confirmation of the hypothesis will require dynamic imaging experiments to demonstrate directly the active recruitment of PML to specific loci following mitosis. Finally, it is not clear whether PML bodies are homogeneous, and to what extent biochemical heterogeneity would affect their function. We have noted, for example, the differential accumulation of PML-body-associated proteins at these bodies following cellular stress in cell lines from different tissue types, as well as in PML bodies within the same cell. In addition, at least nine isoforms of the PML protein have been described, and their contribution to the structure and function of PML bodies is only now being addressed. Although the answer to these outstanding questions might shed light on the possible function of PML bodies in regulating gene activity, we could still find it difficult to distinguish whether the chromatin environment in the vicinity of the PML body dictates its biochemical composition and structure or vice versa.

G.D. is a Senior Postdoctoral Fellow of the Canadian Institutes of Health Research (CIHR). D.P.B.-J. is the recipient of the Canada Research Chair in Molecular and Cellular Imaging. This work was supported by an operating grant from the CIHR (FRN14311) to D.P.B.-J.