The promyelocytic leukemia (PML) protein has been implicated in many cellular pathways, but it is unclear whether the accumulation of PML and other proteins into PML nuclear bodies is a regulated or random process. In this paper we have used a variety of physiological stresses, including heat stress, Cd+2 exposure and adenovirus E1A expression, as tools to study the principles underlying the assembly/disassembly, integrity and dynamic behavior of PML bodies. Using live-cell imaging and immunofluorescence microscopy, we observe that PML bodies are positionally stable over time intervals of a few hours. After stress, however, microstructures form as a result of fission or budding from the surface of `parental' PML bodies. Since new PML bodies do not form at new locations, and the relative sizes observed before heat shock are preserved after recovery, we conclude that there are pre-determined locations for PML bodies, and that they are not random accumulations of protein. Over-expression of small ubiquitin-like modifier (SUMO-1) prevents stress-induced disassembly of PML bodies, implicating SUMO-1 as a key regulator of PML body integrity. Stress-induced fission of SUMO-1-deficient microstructures from parental PML bodies may be a mechanism to change local chromatin domain environments by the dispersal of protein factors. PML bodies may provide a useful paradigm for the dynamics and integrity of other supramolecular protein complexes involved in processes such as transcription, RNA processing DNA repair and replication.
The mammalian eukaryotic nucleus is a highly structured organelle containing a variety of specialized domains that accumulate proteins of specific functions. These large multi-protein domains have been shown to possess dynamic characteristics. At the molecular level, fluorescence recovery after photobleaching experiments indicate that most proteins are highly mobile in the nucleoplasm and their exchange with chromatin or nuclear subdomains is dynamic (Misteli, 2001). At the level of the subdomains themselves, dynamic behavior is also exhibited. For example, nuclear speckles or interchromatin granule clusters undergo fission events involving supramolecular complexes (Misteli et al., 1997; Kruhlak et al., 2000). Cajal bodies move through the nucleoplasm by constrained diffusion, through repeated cycles of association and disassociation with chromatin by an energy-dependent mechanism. These bodies also undergo fission events (Platani et al., 2000; Platani et al., 2002). It has been proposed that promyelocytic leukemia (PML) bodies are based on a supramolecular subunit(s) (Kentsis et al., 2002). This raises the possibility that PML bodies could assemble or disassemble by fusion or fission mechanisms. Finally, a subclass of PML bodies has been reported to have the potential to move through the nucleoplasm as intact entities (Muratani et al., 2002).
PML bodies have been implicated in many cellular pathways, including control of apoptosis (reviewed by Salomoni and Pandolfi, 2002; Eskiw and Bazett-Jones, 2002), viral pathogenicity (reviewed by Regard and Chelbi-Alix, 2001) and control of higher order chromatin structure (de Jong et al., 1996). Several models have been proposed for the function of PML bodies. In one model, PML bodies have no direct function but represent random aggregations of excess nucleoplasmic proteins. In a second model, the bodies may function as storage sites, maintaining proper nucleoplasmic levels of particular factors. The bodies may also serve as sites for post-translational modification of factors. For example, the co-accumulation of the protein acetyltransferase cyclic AMP-response element binding(CREB)-binding protein (CBP) at PML bodies along with one of its substrates, p53, may constitute an important step in regulating its modification (LaMorte et al., 1998; Doucas et al., 1999; von Mikecz et al., 2000; Bandobashi et al., 2001; Boisvert et al., 2001). In a third model, PML bodies may function as sites of specific nuclear events. For example, early gene transcription and replication of several viruses occur in the vicinity of PML bodies (Ishov et al., 1997). Recently, Shiels and co-workers have shown that the MHC class I gene cluster is located on the surface of a subset of PML bodies, indicating that PML bodies may participate in the regulation of particular genes, or gene families (Shiels et al., 2001). Nevertheless, it is still unclear whether PML bodies play an active role in nuclear events, whether they form as a result of nuclear events, or are simply random accumulations of excess nuclear factors (Negorev and Maul, 2001).
In this paper we have used a variety of cellular stresses as tools to study the principles underlying the assembly/disassembly, integrity, size, location and dynamic behavior of PML bodies. Environmental stresses such as heat shock, heavy metal shock, or viral protein expression, cause the dissociation of PML bodies into many smaller punctate domains (Maul et al., 1995), which we will refer to as microstructures. Using live-cell imaging and immunofluorescence microscopy, we observe that PML bodies are positionally stable over time intervals of a few hours. After stress, however, microstructures form as a result of fission or budding from the surface of `parental' PML bodies. Energy depletion does not affect the microstructure diffusion rate. All microstructures eventually fuse with parental PML body remnants. Since new PML bodies do not form at new locations, and the relative sizes observed before heat shock are preserved after recovery, we conclude that there are pre-determined locations for PML bodies, and that they are not random accumulations of protein. Finally, small ubiquitin-like modifier (SUMO-1) is not detected in PML-containing microstructures immediately following stress, but accumulates in these structures during recovery. Furthermore, transient over-expression of SUMO-1 prevents the stress-induced dissociation of PML into microstructures. Together, these data indicate a key role for SUMO-1 in the mechanism of stress-induced dissociation and subsequent re-assembly of PML bodies.
Materials and Methods
Cell culture, transfection and protein extracts
SK-N-SH cells were cultured on glass cover slips in JMEM or DMEM medium to high confluency (80-90%). Cells were transfected with either 2 μg of pCMX-GFP-PML3 DNA (PML IV) (Jensen et al., 2001) [a gift from R. Evans, The Salk Institute for Biological Studies, La Jolla CA; described previously by Boisvert et al. (Boisvert et al., 2001)], DSred-PML IV DNA (a gift from M. Hendzel, University of Alberta), eGFP-SUMO1 (a gift from Y. Kim, National Institutes for Health) or pEGFP-E1A (see below) per cover slip using 5μl lipofectamine 2000 (Invitrogen) in 200 μl Opti-MEM. The DNA/lipofectamine mix was placed on the cover slips and incubated in 2 ml Opti-MEM (BRL-Gibco) for 6 hours at 37°C. Opti-MEM was replaced with JMEM or DMEM and the cells were allowed to express the transgene for 12 to 24 hours at 37°C. pEGFP-E1A expresses the 289-residue Ad5 E1A protein and was constructed by subcloning the full-length E1A cDNA into the expression vector pEGFP-C2 (Clontech Laboratories, Palo Alto CA). HEp-2 and U2OS cells stably expressing GFP-PML I or GFP-PML IV (a gift from J. Taylor, Medical College of Wisconsin) were cultured in 0.5 mg/ml or 1.5 mg/ml G418, respectively, in DMEM. GFP-PML I/IV stable cells produce PML bodies in size, composition (e.g. Sp100) and number comparable to that of naive cells. Infections of this cell line with Herpes virus or transfection with ICP0 leads to the complete disruption of PML bodies, indicating that the levels of PML expression are physiologically relevant (J. Taylor, personal communication).
Energy depletion studies were conducted by incubating SK-N-SH cells growing on cover slips in JMEM or DMEM with 50 μM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP; Sigma, catalogue no. C-2920) and 200 mM 2′-deoxy-glucose (Sigma, D-3179) for 2 hours.
Protein extracts were prepared from 80% confluent cultures of cells grown in 10 cm culture dishes. Heat shock was performed at 43°C for 30 minutes and cadmium shock with 50 μM Cd+2 for 2 hours. Cells were then lysed in 800 μl of 6 M guanidinium-HCl. Proteins were precipitated by mixing with 800 μl of 10% trichloroacetic acid, briefly vortexed, followed by a 20-minute incubation on ice. The mixture was centrifuged at 4°C for 20 minutes, and the resulting pellet was washed with 400 μl of 95% ethanol, dried and resuspended in 400 μl SDS-PAGE sample buffer. For western blots, equal volumes of sample were loaded per lane, and the relative proteins concentration determined by Coomassie Blue staining. Levels of HSP70 protein were analyzed using mouse anti-HSP70 (Stressgen).
Immunofluorescence microscopy, cell imaging and quantification
Cells grown on glass cover slips were fixed in 4% paraformaldehyde/PBS (pH 7.5) for 5 minutes at room temperature, permeabilized in 0.5% Triton X-100 in PBS for 5 minutes at room temperature. PML protein was visualized with the monoclonal antibody 5E10 (a gift from R. van Driel) or polyclonal rabbit anti-PML (catalogue no. AB-1370, Chemicon International). Sp100 was labeled with rabbit anti-Sp100 antibodies (catalogue no. AB1380, Chemicon International) and CBP was labeled with rabbit anti-amino-terminal CBP antibodies (catalogue 3 sc-369, Santa Cruz Biotechnology, Inc.). Endogenous SUMO-1 localization was determined by labeling cells with mouse anti-GMP-1 (catalogue no. 33-2400, Zymed Laboratories).
Nascent transcripts were visualized by the incorporation of 5′-fluorouridine (Sigma Biotechnologies) followed by labeling cells with mouse anti-BrdU primary antibodies (Sigma Biotechnologies). Secondary labeling was conducted using goat anti-mouse Cy3, goat anti-rabbit Cy5, goat anti-rabbit A488 or donkey anti-goat Cy5. Cover slips were mounted on glass slides using PBS/90% glycerol containing 1 mg/ml paraphenlenediamine (PPDA) and 1 μg/ml of the DNA specific stain, DAPI. Images were collected on a Leica Microsystems DMRA2 upright microscope. Haze removal by deconvolution was performed using either Open Lab 3.0.7 software (Improvision) or Volocity 2.0 software (Improvision). Images were false-colored using Image J (NIH) software and presented using Adobe PhotoShop 6.0. Quantification of relative protein composition of PML-containing structures was obtained from unprocessed optical sections of fluorescence images. Rectangular regions-of-interest masks were drawn around structures. Integrated intensities were obtained from mean values above background multiplied by the area of the region-of-interest. The same mask was used in both the PML's and the other protein's fluorescence images. The background value was obtained by masks drawn in the immediate vicinity surrounding the structures of interest.
For analysis of live cells expressing GFP-PML, cover slips were placed on a live cell environmental chamber constructed in our laboratory. The sealed chamber accommodated 6 ml of CO2-equilibrated JMEM or DMEM medium; which maintains the cells in a buffered environment at 37°C during the course of an experiment. For experiments involving transient transfection, cells expressing low levels of GFP-PML IV were selected, imaged, and the locations of these cells were marked on the cover slips with a high resolution grid. To heat shock the cells the chamber was placed in a 43°C water bath for 30 minutes and then returned to the microscope for imaging, or shocked directly on the stage using a stage heater (Leica). This procedure was duplicated in experiments using HEp-2 or U2OS cells stably expressing GFP-PML I/IV. 3-D data sets were collected with the upright microscope previously described but fitted with a Wave FX spinning disk confocal unit (Quorum Technologies, Inc.) utilizing a Yokagawa dual disk confocal head. We used the 488 nm line of a multi-line argon laser and an excitation/emission filter wheel appropriate for GFP.
Integrated intensities of PML and SUMO-1 signals were measured for PML bodies, PML remnants and microstructures. Integrated intensity values were defined as (mean immunofluorescence signal intensity measured over the structure minus the mean intensity in the surrounding background) multiplied by the area of the structure. If the average SUMO-1 signal over a microstructure that contains the lowest SUMO-1 levels measured in a particular cell was less than the surrounding background region, the resulting negative signal was used to normalize the SUMO-1 signals in that cell, so that a value of 1 was obtained for the mean signal intensity of that microstructure. This precluded calculating an infinite ratio of PML:SUMO-1.
Correlative fluorescence microscopy and transmission electron microscopy
Following immunolabeling, SK-N-SH cells were post-fixed for 5 minutes in 2% gluteraldehyde/PBS (pH 7.5), followed by dehydration in consecutive graded steps of ethanol (30%, 50%, 70%, 95% and absolute ethanol) at room temperature for 15 minutes at each step. Cells were then embedded in Quetol 651 (Electron Microscopy Sciences) following the instructions supplied with the resin. Small blocks of Quetol 651 containing cells previously identified and marked (Ren et al., 2003), were glued to bullets and sectioned to the desired thickness using an ultramicrotome (Leica Microsystems). Serial sections were collected on 400 mesh copper grids (Electron Microscopy Sciences). Sections on grids were viewed with the fluorescence microscope and images of the regions of interest of specific cells were collected. These regions of the section were then imaged at the ultrastructural level at 200 kV in the transmission electron microscope (Tecnai 20, FEI). Energy filtered images were collected using a post-column imaging filter (Gatan) as described elsewhere (Bazett-Jones and Hendzel, 1999; Boisvert et al., 2000).
Heat stress, heavy metal exposure and expression of adenovirus E1A causes PML bodies to partially disassemble to form PML-containing microstructures
Cellular stresses can lead to a decrease in the number and size of PML bodies and the formation of smaller PML-containing complexes that we refer to as `microstructures'. For example, HEp-2 cells treated with Cd+2 ions produce microstructures (Maul et al., 1995; Nefkens et al., 2003). Microstructures are also observed when a variety of viral proteins are expressed, including the Herpes Simplex early viral protein ICPO (Maul et al., 1995). To study the formation of these structures in live cells, we first determined the conditions under which microstructures would form. We stressed SK-N-SH and U2OS cells by either heat shock at 42°C for 1 hour, exposure to Cd+2 at 50 μM for 2 hours, or 12 hours of adenovirus type 5 E1A protein expression (Fig. 1A). Under all of these conditions, microstructures were observed by immunofluorescence. This is significant because diverse conditions, such as environmental and viral protein-induced stress, lead to the same morphological effect on PML bodies. Under heat shock conditions, stably transfected U2OS or HEp-2 cells expressing GFP-PML I (882 aa) or GFP-PML IV (633 aa) [PML nomenclature described by Jensen et al. (Jensen et al., 2001)], or transiently transfected SK-N-SH cells expressing GFP-PML IV or DSred-PML IV were initially performed at 42°C for 1 hour. Over 90% of cells that express normal levels of all PML isoforms will form microstructures. In contrast, only approximately 5% of transfected cells generate microstructures under these heat stress conditions. We found, however, that a heat shock at 43°C for 30 minutes was sufficient to generate microstructures in >90% of cells transiently or stably expressing exogenous copies of PML I or PML IV. Changes in the expression pattern of HSP 70 detected by immunofluorescence (Fig. 1B), as well as increased levels, determined by western blot analysis of protein extracts from cells that were heat shocked with this protocol (Fig. 1C) indicated a typical heat stress response (Morimoto, 1993). Moreover, these heat shock conditions did not affect cell survival, since approximately 95% of cells survived the treatment (data not shown).
Heat stress was more suited than cadmium exposure or E1A expression for studying the dynamic properties of microstructures mainly because heat stress is immediately reversible. Although cadmium stress is reversible, the precise timing of recovery is indeterminate because of the unknown time required to wash Cd+2 out of cells. With heat stress, recovery begins immediately following a return to normal growth conditions.
Microstructures are generated from parental PML bodies
We wished to determine the origin of the PML-containing microstructures. One possibility is that they arise directly from `parental' PML bodies. Alternatively, the microstructures may form by accumulation or aggregation of molecular PML from the nucleoplasm. We measured the GFP signal intensities in single unprocessed 2-D images of a PML body before and after heat shock. For more than 30 PML bodies measured, we observed that the GFP-PML signal intensity dropped dramatically. In cells either transiently expressing GFP-PML IV (data not shown) or stably expressing GFP-PML I (Fig. 2A), we observed that PML bodies lost from 30% to 70% of their GFP signal intensity, with an average loss of approximately 50%. The nucleoplasmic background measured over the entire nuclear area outside PML body remnants, which includes signal from microstructures, increased by approximately 5%. This is consistent with the loss of signal from the bodies and indicates that the GFP fluorescence is not labile to exposure to the higher temperature. As cells recovered from the stress, the intensity of the GFP signal in PML bodies increased.
To further characterize the origins of PML containing-microstructures, we collected 3-D data sets over time of U2OS cells stably expressing GFP-PML IV following heat shock at 43°C on the microscope stage. Image collection was initiated 20 minutes after the induction of the stress stimulus. (Fig. 2B; Movie 1, http://jcs.biologists.org/supplemental/). In this example, the PML body marked with an arrow (in I) appears as a round structure. 40 seconds later (II) a bleb appears on the surface and persists at this location for approximately 100 seconds (III), until it is released to the surrounding nucleoplasm (IV). Examination of this event from different angles reveals that this structure does bud from the PML body itself and does not arise from aggregation of PML from the nucleoplasm. Similar results were obtained when GFP-PML IV was expressed transiently, or when GFP-PML I and IV were stably expressed in either U2OS or HEp-2 cells.
Microstructures are mobile and fuse during recovery
We then analyzed the movement and fate of both PML-containing microstructures and PML body remnants following cell stress. Using 2-D and 3-D time-lapse imaging of GFP-PML, we observed that microstructures were mobile entities with dynamic characteristics consistent with constrained diffusion (see Movies 2 and 3, http://jcs.biologists.org/supplemental/). The movement of these structures, however, was not uniform over time. In the example shown in Fig. 3A,B, a microstructure moves in relatively short distances between each time point while in region I, then migrates to region II where it can move longer distances between time points than it could in region I. Finally, it migrates to region III where again it is restricted to smaller distances between each time point. Our interpretation of this trajectory over time is that the microstructure is confined to a nucleoplasmic compartment in which its mobility is constrained (region I), then it escapes into a more open compartment in which it can move long distances between time points (region II). Finally, the microstructure leaves this compartment, but enters a third compartment (region III) where once again its mobility is more restricted. A plot of the cumulative distance traveled as a function of time further describes the mobility of this microstructure (Fig. 3C). It is important to note that in the 648 seconds of this analysis, most microstructures are characteristic of those in region I, and only a small fraction of the microstructures display the mobility characteristic of region II. Over longer periods of observation (e.g. 30 minutes), however, all microstructures are able to escape such constrained regions and become more mobile for a period of time. In contrast, PML body remnants remain positionally stable.
To confirm the broader relevance of microstructure dynamics occurring with heat stress, we observed cells expressing GFP-PML IV that were treated instead with 50 μM Cd+2, and cells expressing DSred-PML IV that were stressed by GFP-E1A expression. Under these conditions, the resulting microstructures displayed identical characteristics to those in heat stressed cells (Fig. 3A,B). Similarly, PML body remnants remained positionally stable following Cd+2 exposure or E1A expression (Movies 4 and 5, http://jcs.biologists.org/supplemental/).
To compare the motion of PML-containing microstructures with the movement of other nuclear components, including Cajal bodies and chromatin, we collected 2-D data sets at 1 second intervals. Analysis of the mean squared displacement <d2> (MSD) (Chubb et al., 2002; Platani et al., 2002) revealed that there were 2 subsets of PML-containing microstructures (Fig. 3D). The first appeared to move slowly, with an average diffusion coefficient of 2.9×10–4 μm2/second, comparable to that of chromatin (Chubb et al., 2002). We conclude later that the mobility of such microstructures is dictated by the inter-chromatin domain in which they are trapped, and corresponds to the type of movement of a microstructure in regions I and III (Fig. 3B,C). The average diffusion coefficient of a second subset was 1.4×10–3 μm2/second, though some had instantaneous diffusion constants as high as 5×10–2 μm2/second. These microstructures are less constrained by local chromatin domains, corresponding to microstructures in a nuclear domain-like region II (Fig. 3B,C). Whereas constrained microstructures move with velocities averaging 50-70 nm/second, the free-moving microstructures can move at rates up to 0.5 μm in a 1 second interval (Fig. S1, http://jcs.biologists.org/supplemental/).
To better understand the physical constraints that affect the microstructures' dynamic behavior, we obtained ultrastructural information by correlative fluorescence microscopy and analytical electron microscopy (Boisvert et al., 2000; Ren et al., 2003). Heat shocked SK-N-SH cells were labeled for PML protein and prepared for electron microscopy. Physical sections 60 nm in thickness were first imaged by fluorescence microscopy to identify PML body remnants and microstructures (Fig. 4A). Nitrogen maps (green, Fig. 4B) were used to identify the PML body remnants seen in the fluorescence images of the same serial section. Microstructures such as the one labeled `a' could be found in the electron microscope images by comparing the fluorescence images with low and intermediate magnification nitrogen maps and applying triangulation techniques. Microstructures were defined as PML-containing structures that were present in a maximum of two physical sections, thus distinguishing them from grazing sections of PML body remnants. Their morphology and lack of phosphorus frequently make then difficult to distinguish from the protein-based architecture of the nucleoplasm. Microstructure `a' (Fig. 4C), however, was identified by triangulation measurements and could be resolved from other near-by protein-based structures. Phosphorus maps (red) were used to distinguish nucleic acid from protein-based structures. We observed that microstructures are protein-based entities, approximately 40-100 nm in size, containing no detectable levels of phosphorus above nucleoplasmic background. Some of these microstructures occupy large open regions in the nucleoplasm that are devoid of chromatin (outlined in the P map in Fig. 4C). Microstructure mobility in such regions would probably be described by the mobility characterized by region II (Fig. 3), corresponding to the subset of microstructures with the higher diffusion constant (Fig. 3D). The dimensions of microstructures are similar to the diameters of many of the much narrower `channels' and spaces between chromatin fibers throughout the nucleoplasm (indicated by * in Fig. 4C). A microstructure in such regions of the nucleoplasm would appear relatively constrained in the live-cell imaging experiment (region I, Fig. 3), diffusing at the same rate as chromatin.
PML-containing microstructures fuse during recovery but do not form new PML bodies
We then followed the fate of microstructures over long time intervals as cells recover from stress. 3-D data sets were collected and projected into a single plane (Fig. 5A,B). Thus, every PML body remnant and microstructure could be accounted for over the time course. Images were collected before heat shock, immediately after heat shock, and at 5-minute intervals for a further 30 minutes. A slight change in nuclear shape, based on the relative position of the PML bodies and remnants, was observed between the pre- and post-heat shock time points, but no further shape changes were observed over the next 30 minutes. We observed that the microstructures were able to move hundreds of nanometers within the 5-minute interval (rectangle, Fig. 5A,B). Furthermore, microstructures were observed to fuse with each other as they moved through the nucleoplasm (indicated by arrowheads in Fig. 5B). Quantification of GFP signal also confirmed that the microstructures were fusing with each other since the intensity of the structures relative to the background increased over the time course. This is demonstrated by three microstructures (indicated in Fig. 5A); the mean signal of the structures labeled 1, 2 and 3, increased 5.7, 2.5 and 3.4 fold above background, respectively, between the post-heat shock and the 30 minute time points. We conclude that the increase in the mean signal intensity occurs as microstructures fuse to produce larger microstructures during the recovery period. To determine whether the mobility of the microstructures is based on diffusion or on an energy-dependent molecular motor mechanism, we observed microstructure movement during the recovery phase following heat shock in cells that were treated with FCCP and 2′-deoxy-D-glucose to deplete ATP. Under these conditions, microstructures were as mobile as those in cells under normal conditions (data not shown), supporting our conclusions that the mobility of the microstructures is an example of constrained diffusion.
We also observed that microstructures fused with the larger positionally stable PML body remnants. Using 3-D reconstructions of images of U2OS cells stably expressing GFP-PML IV at 30 minutes post-heat shock, we observed microstructures merging or fusing into PML body remnants (Fig. 5C; Movie 6, http://jcs.biologists.org/supplemental/). In the example shown, a microstructure contacts a PML body remnant, and fuses with the remnant over a 13-second time interval. To achieve better temporal resolution of these events, we captured 2-D images at 30 minutes after heat shock (data not shown). We carefully monitored 25 such fusion events in 22 different cells. In all cases, no microstructure was seen after the apparent fusion event. We conclude that once this fusion has occurred, it is irreversible and that the fusion of microstructures to each other and ultimately to larger PML body remnants is the primary mechanism for the transition of PML body remnants back to mature and fully functional PML bodies.
If PML bodies were random accumulations of PML and other proteins, we would expect that following heat shock and subsequent recovery, the liberated PML microstructures might nucleate and create new PML bodies at new nuclear locations. To address this possibility, we examined more than 40 transient and stable GFP-PML-expressing cells prior to, and up to 4 hours following, heat shock. We observed that PML bodies were positionally stable over the entire time course (Fig. 5D), after accounting for minor global shape changes of the nucleus. Importantly, no new PML bodies were observed. A random aggregation of PML liberated from `parental' PML bodies does not occur. Since approximately 50% of the GFP-PML signal is lost from the parental bodies, a cell with 10 PML bodies prior to heat shock would be expected to potentially generate 5 new PML bodies after recovery, if the formation was driven by a random aggregation of excess nuclear protein. More importantly, the rank order of the relative sizes of the PML bodies after 3 hours recovery is frequently the same as that before heat shock. Not only is the nuclear position of the PML body determined, but the size also appears to be pre-determined. To illustrate this, the rank size order, based on the integrated intensity of the GFP signal in a cell stably expressing GFP-PML, was obtained for 11 PML bodies as a function of time following the stress. Immediately following heat stress (Fig. 5D, postHS), the rank order was scrambled. Following 3 hours of recovery, however, the original size order was essentially restored, with the exception that the eighth largest remnant did not fully recover, dropping to the tenth position in the rank order (Fig. 5D). In 16 of 19 such cells, the rank order was completely preserved following recovery. In each of three cells, a pair of PML bodies exchanged positions in their rank order of size. No cells were observed where more than a single pair of bodies underwent a rank order inversion. We conclude that both the size and location of PML bodies may represent a regulated event, and that PML bodies are not random accumulations of PML and other associated proteins.
Biochemical composition of microstructures indicates a role for SUMO-1 modifications in PML body stability
We wished to relate the structure and dynamic behavior of PML-containing entities before stress and during recovery to their biochemical composition. Immunofluorescence microscopy was used to measure subtle biochemical differences between the various PML-containing structures within single cells. This approach provided the spatial resolution and hence the body-to-body variability that cannot be obtained by conventional biochemical techniques that rely on extraction protocols (e.g. western blot analysis). SUMO-1 has been found to be covalently attached to PML protein and has been implicated in PML body stability (Muller et al., 1997; Sternsdorff et al., 1997; Goodson et al., 2001). We found that the PML body remnants maintained high levels of SUMO-1 immediately following both heat and Cd+2 stress, whereas the microstructures had no detectable levels of SUMO-1 accumulation (Fig. 6A,B). Line scans of PML and SUMO-1 signals across PML body remnants and microstructures (Fig. S2, http://jcs.biologists.org/supplemental/) clearly showed that quantification of the relative PML and SUMO-1 levels as detected by immunofluorescence was feasible (Tables 1, 2). Similarly, microstructures formed by E1A expression contained no detectable SUMO-1 (not shown). The presence of SUMO-1 in the remnants indicates that either PML and/or other components are modified by SUMO-1 following heat stress. After PML, the most prevalent SUMO-1 modified protein in PML bodies is Sp100. Since most PML body remnants (and microstructures) contained little or no Sp100 (not shown), the likely target for SUMO-1 modification is PML protein itself. Furthermore, SUMO-1/PML levels remain high in the PML body remnants, yet SUMO-1 is undetected in the microstructures immediately following the stress.
We then asked whether SUMO-1 modifications are involved in the re-assembly of PML bodies through the observed fusion of microstructures during the stress recovery phase. In contrast to observations immediately following stress, at 2 hours (not shown) and 4 hours post-stress microstructures contained high levels of SUMO-1 (Fig. 6A, Table 1). We conclude that SUMO-1 plays a role in the re-assembly fusion mechanism of PML body recovery. Further support for SUMO-1 involvement in PML body stability and assembly was obtained with cells over-expressing GFP-SUMO-1. We observed that GFP-SUMO-1 expression prevented the formation of PML-containing microstructures in SK-N-SH cells that were heat shocked at 43°C for 30 minutes, or stressed by 50 μM Cd+2 for 2 hours (Fig. 7). Protection against microstructure formation probably occurs by a shift in the equilibrium of PML modification towards the SUMOylated form. Since increased levels of SUMO-1 prevent stress-induced disassembly of PML bodies, we conclude that SUMO-1, probably through PML, or through another unidentified SUMO-modified protein, plays a key role in regulating PML body integrity. The mechanism by which SUMOylation maintains this integrity, and the mechanisms involved in modulating SUMOylation levels in response to cellular events such as heat shock, remain to be elucidated.
Evidence for supramolecular assembly/disassembly of PML bodies
Following periods of cellular stress, global levels of transcription are greatly reduced, and PML bodies are disrupted, giving rise to microstructures in the nucleoplasm (Fig. 1). Although we cannot discount that molecular PML may be released from sites other than PML bodies during stress, we have observed PML containing microstructures `budding' from parental bodies (Fig. 2). These fission events indicate that PML found in PML microstructures originates primarily from PML bodies rather than the nucleoplasm. The newly formed microstructures lack Sp100 and SUMO-1, are mobile within the nucleoplasm, and move up to 0.5 μm in a 1-second interval, whereas PML body remnants and PML bodies remain positionally stable over extended periods of time (Fig. 3).
During the recovery phase, microstructures collide with each other and have the propensity to fuse, creating larger microstructures. These, in turn, fuse directly with the PML body remnants (Fig. 5). This behavior indicates that PML bodies are not a uniform, homogeneous `polymer'. Instead, PML bodies may be composed of units or modules that are linked together as supramolecular assemblies, a characteristic of proteins containing the RING domain, such as PML (Kentsis et al., 2002). During the recovery phase, microstructures do not fuse with each in order to create new PML bodies in new locations. Moreover, after significant loss of PML protein from the bodies as a result of stress, an underlying mechanism dictates the size of each body, since the relative sizes of the bodies are conserved between the pre-stressed state and after recovery (Fig. 5). The physico-chemical basis of PML body size conservation could be a passive constrained diffusion mechanism and/or an active recruitment mechanism of PML body components in response to some underlying biological function, such as transcriptional regulation. In the passive constrained diffusion model, molecular and/or supra-molecular PML microstructures released during stress are confined to a particular domain or region of the nucleus by local chromatin organization, and periodically fuse to form larger microstructures. When these microstructures are permissive for fusion with parental PML bodies (perhaps involving changes in SUMOylation levels), they fuse by constrained diffusion with the nearest PML body within that region of the nucleus. Thus, PML body size during recovery is a function of (1) the local molecular or supra-molecular concentration of PML body components found within a particular region or domain of the nucleus, and (2) the size of the inter-chromatin space in which the parental PML body is found.
The second model predicts that the size and positional conservation of PML bodies are based on an active mechanism, dictated by the underlying biological function(s) of the PML body remnant after stress recovery, such as the transcriptional regulation of the chromatin found in its vicinity (Boisvert et al., 2000). We have observed that PML microstructures can move large distances within the nucleus. Therefore, it is possible that PML body components could redistribute throughout the entire nuclear volume during stress recovery. The consequence would be a lack of conservation of PML body size. Regardless of whether an active or a passive recruitment process is involved, it is clear that the formation of PML bodies is not based on a random aggregation of excess PML protein. If PML bodies act only as storage sites or sites of post-translational modifications of body components, such as CBP and p53 (Fogal et al., 2000), then the relative size of PML bodies would not be expected to be conserved, nor show positional stability. Because our observations support conservation of size and position, we favor a recruitment mechanism that reflects the activity of the specific chromatin domains that surround each body. For example, the recruitment of transcriptional co-activators, such as CBP (LaMorte et al., 1998) may create a domain enriched in acetylated chromatin and other factors, thus creating a favorable environment for transcriptional regulation (Boisvert et al., 2001). We suggest that the pre-determined locations reflect an underlying protein-based architecture that may arise within an inter-chromatin domain in response to one or more of the putative functions of the PML body.
It appears that the presence of SUMO-1, is responsible for PML body integrity. PML proteins with mutations that remove the SUMOylated residues are targeted to the nucleus but not to PML bodies (Zhong et al., 2000). A decrease in the amount of SUMO-1 in PML body remnants was observed immediately following heat shock, and no detectable SUMO-1 was associated with microstructures following heat stress, Cd+2 exposure or with E1A expression (Fig. 6). During recovery, however, levels of SUMO-1 increased within microstructures, corresponding to the time when fusion events were frequently observed. Further support for a role for SUMO-1 in stabilizing PML body structure was obtained when GFP-SUMO-1 was transiently over-expressed. Increased levels of SUMO-1 prevented the formation of microstructures in cells heat shocked at 43°C for 30 minutes or in cells stressed by 50 μM Cd+2 (Fig. 7). We suggest that the mechanism of stress-induced disassembly and subsequent re-formation of PML bodies involves a shift in the equilibrium of SUMO-1 modification of PML and/or other PML components. This is consistent with the involvement of a SUMO protease in PML body composition and integrity (Best et al., 2002). It has been reported that SUMO-1 may be cleaved from PML protein once it is incorporated into PML bodies (Maul et al., 2000). Since we observed SUMO-1 localization in PML body remnants that had dramatically decreased levels of Sp100, we propose that PML is still modified by SUMO-1 after PML deposition at PML bodies. However, it remains to be determined whether the loss of SUMO-1 causes the release of PML protein from PML bodies, or if the release of PML protein results in de-SUMOylation.
PML-containing microstructures move by constrained diffusion in the nucleoplasm
Some of the dynamic characteristics of PML bodies are similar to those of other nuclear sub-domains. For example, we have demonstrated that under stress conditions PML bodies undergo fission/blebbing and fusion/rejoining events, which occurs with both interchromatin granule clusters (IGCs) (Misteli et al., 1997; Kruhlak et al., 2000) and Cajal bodies (Platani et al., 2000). Whereas IGCs and PML bodies appear to be positionally stable, Cajal bodies move with a velocity of approximately 10-15 nm/second (Platani et al., 2000). However, in this study we identified 2 subsets of PML-containing microstructures that can be distinguished with respect to mobility. Our interpretation is that one subset of microstructures are trapped within pockets of chromatin and, therefore, their diffusion constants are similar to that of mobile chromatin domains (Marshall et al., 1997; Chubb et al., 2002). In these domains, the microstructures exhibit an average velocity of 50-70 nm/second and a diffusion constant of 2.9×10–4 μm2/second. From the physical dimensions of microstructures measured by ESI (40-100 nm; Fig. 4), we predict a molecular mass of 0.5-5 MDa. This places them in the same size range of many multi-subunit molecular complexes such as the basal transcription apparatus, chromatin remodeling complexes and spliceosomes. We suggest that the motion of PML-containing microstructures may represent a paradigm for the dynamic behavior of other large complexes in intranuclear trafficking. The second subset of microstructures may occupy large, chromatin-free channels which allows for diffusion constants that are approximately an order of magnitude greater than those in the first subset; such microstructures have been observed to move 0.5 μm in 1 second. The energy-independent movement of microstructures is consistent with a lack of functional contacts with other nuclear structures, in contrast to the Cajal body, whose movement is thought to be constrained by functional interactions with chromatin (Platani et al., 2000; Platani et al., 2002). The structures can move very rapidly until they become trapped within nuclear sub-domains. Frequently, microstructures `escape' through openings within the nucleoplasm and are able to travel large distances in short time intervals, before once again becoming trapped, probably within pockets of chromatin. Since the movement of PML-containing microstructures is energy-independent, the mobility is characteristic of constrained diffusion.
Chromatin dynamics is an important consideration in understanding nuclear function (Chubb and Bickmore, 2003; Marshall, 2002). In addition, however, we propose that the dynamic behaviour of specialized protein-based compartments, such as the PML body, may also have consequences for nuclear function. These are not static entities; they are capable of assembly and disassembly in response to physiological stimuli, and have characteristic mobility properties in the nucleoplasm. Moreover, an interplay may exist between these dynamic protein structures and the surrounding chromatin, such that the assembly and disassembly of a body may influence, and/or be influenced by, the local chromatin organization (Boisvert et al., 2000). There are already precedents for such interplay as illustrated by the changes in chromatin mobility observed by Chubb et al. (Chubb et al., 2002) in response to the disassembly and reformation of the nucleolus during reversible transcriptional inhibition. We suggest that the dynamic behavior of PML bodies reported here serves both as a model for other nuclear bodies, such as Cajal bodies (Platani et al., 2002), and for large supramolecular complexes with specialized functions, including transcription initiation, chromatin remodeling, DNA repair, recombination and replication.
We thank Ying Ren and Ren Li for excellent technical assistance. We thank Dr Roel van Driel for the gift of the 5E10 PML antibody, Dr R. Evans for the PML expression vector, and Dr Dean Jackson for the protocol for labeling nascent transcripts with 5′-fluorouridine. We thank Dr J. Taylor for the stable cell lines expressing PML I and IV and for providing characterization information. We also thank Dr J. R. Chubb for advice on the mean squared displacement analysis. We would especially like to thank Mr Reg Sidhu and Dr Robin Battye (Quorum Technologies Inc.) for the assistance and use of the Wave FX spinning disk confocal unit in the collection of 3-D dynamic data. G.D. is a Senior Postdoctoral Fellow of the Canadian Institutes of Health Research. J.S.M. is a Scholar of the Canadian Institutes of Health Research. The research was funded by an operating grant from The Cancer Research Society, Inc. D.P.B.-J. is the recipient of a Canada Research Chair in Molecular and Cellular Imaging.
Movies and supplemental data available online
- Accepted July 8, 2003.
- © The Company of Biologists Limited 2003