Heat shock and Cd²⁺ exposure regulate PML and Daxx release from ND10 by independent mechanisms that modify the induction of heat-shock proteins 70 and 25 differently

ND10, also called PML bodies, are nuclear protein accumulations associated with the nuclear matrix that have been intensely studied because of their apparent involvement in various diseases(Seeler and Dejean, 1999) and viral infections (Maul, 1998). In response to global challenges, levels of ND10-associated proteins either increase by recruitment into ND10 or decrease by release of proteins from this site. Infection or inflammation leads to the accumulation of ND10-associated proteins through upregulation by interferon (see review by Seeler and Dejean, 1999),whereas stress disperses these proteins(Maul et al., 1995),suggesting the involvement of ND10 in two distinct cellular defense mechanisms.

The nucleus of eukaryotic cells is a complex and dynamic structure with different domains involved in specific processes. For ND10, physiological functions remain largely unknown, although a depot function has been suggested(Maul, 1998; Negorev and Maul, 2001). The first ND10 constituent protein molecularly characterized was Sp100, an autoantigen in primary biliary cirrhosis, which is thought to be a transcriptional repressor (Sternsdorf et al., 1999; Szostecki et al.,1990). Another major component of ND10 is PML, a fusion partner of the RARα (retinoic acid receptor-α) in the t(15;17) translocation from patients with acute promyelocytic leukemia (APL). PML has been suggested to be a tumor suppressor and transcriptional regulator, and might function in apoptosis (Zhong et al.,2000b). PML and Sp100 are both covalently modified by the ubiquitin-like protein SUMO-1 (sentrin-1), which is thought to regulate the localization and/or specific protein interactions of the modified proteins(Kretz-Remy and Tanguay,1999). Other proteins including Daxx, BLM helicase, topoisomerase 3, PAX 3 and CBP also colocalize in ND10 (reviewed by Negorev and Maul, 2001),raising the possibility of repressive or activating transcriptional functions at this site (for a review, see Zhong et al., 2000b). However, no newly synthesized RNA(Ishov and Maul, 1996; Grande et al., 1996) or basal transcription factors have been reported in ND10.

PML plays a key role in the integrity of ND10, because all other ND10-associated proteins are dispersed in the absence of PML(Ishov et al., 1999; Lallemand-Breitenbach et al.,2001; Zhong et al.,2000a). However, the possibility of a proto-ND10 at evolutionarily older eIF-4E sites has been suggested(Cohen et al., 2001). In the present model of ND10 formation and protein recruitment, PML represents the main scaffold protein to which other proteins such as Daxx bind in a SUMO-1-dependent way. Under physiological conditions, this is a dynamic process that might regulate the freely mobile availability of certain components in the nucleoplasm through a regulatable recruitment-release process. Thus, recruitment might be favored on at least two different levels:by increasing the amount of the scaffold protein PML or by enhancing the post-translational sumolation of PML. Indeed, both interferon (IFN) treatment to upregulate PML transcription(Chelbi-Alix et al., 1995; Chelbi-Alix et al., 1998; Fabunmi et al., 2001; Grotzinger et al., 1996; Lavau et al., 1995; Stadler et al., 1995) and overproduction of SUMO-1 enhance the depot function of ND10 by increasing the capacity to recruit proteins (Ishov et al., 1999). However, the mechanism(s) controlling the release has not been identified. The stress-induced dispersion of ND10-associated proteins(Maul et al., 1995) thus provides the concept and assay to probe for a regulated release mechanism.

Various forms of stress can have profound effects on the physiology of the organism or cells. Extracellular environmental insults (hyperthermia,ultraviolet radiation, chemical shock, inflammatory cytokines, heavy metals)are transduced from the cell surface to the nucleus through stress-activated signaling pathways, with those involving the mitogen-activated-protein kinases(MAPKs) being the most conserved. The three major MAPK subfamilies are represented by the extracellular-regulated kinase (ERK), c-Jun N-terminal kinase (JNK) and the stress-activated p38 MAPK (SAPK2), all of which induce the expression of heat-shock proteins (Hsps). Hsps are involved in protecting damaged proteins from subsequent and/or higher stress levels(Morimoto and Santoro, 1998). Upon activation, the Hsp-specific transcription factor HSF1-1 relocates within seconds into 6-10 nuclear stress granules(Jolly et al., 1999), which do not localize with ND10 (Cotto et al.,1997). Pre-existing proteins can be directly activated through one of these cascades, as in the case of Hsp27, which is phosphorylated within 20 minutes of heat shock by MAPKAP kinase-2, a substrate of the p38 MAPK(Adler et al., 1995; Landry et al., 1991). The availability of specific inhibitors permits analysis of the role of such pathways in the release of ND10-associated proteins upon stress(Davies et al., 2000).

A physiological dispersion of ND10-associated proteins occurs during mitosis (Ascoli and Maul, 1991; Everett et al., 1998),although PML is retained in a few aggregates throughout the mitotic phases(Maul and Everett, 1994). The cytoplasmic PML aggregates apparent in the G₁ phase are remnants of ND10 that were excluded from the nucleus during karyogenesis. Everett et al.(Everett et al., 1999) described the appearance of a mitosis-specific,phosphorylated PML isoform and the disappearance of the SUMO-1-modified isoforms during mitosis, providing some indication that protein modifications might regulate the release of ND10-associated proteins.

In the present study, we analyzed the effect of two environmental stress factors, hyperthermia and Cd²⁺ exposure, on ND10 protein retention,and the mechanisms controlling the release of proteins from these nuclear sites. Hyperthermia and Cd²⁺ also trigger reactivation of latent herpes simplex virus type 1 (Fawl and Roizman, 1993; Sawtell and Thompson, 1992). Because herpesviruses use ND10 to begin transcription and replication during lytic infection(Maul et al., 1996; Maul, 1998; Ishov et al.,1996; Ishov et al., 1997) and have evolved proteins that disperse ND10-associated proteins (Maul, 1993; Everett and Maul, 1994),elucidation of mechanisms underlying ND10 protein release and the physiological consequences of this specific release is relevant to broad areas of cell physiology.

Mouse embryonic fibroblasts (MEFs), embryonic stem (ES) cells and their respective PML or Daxx knockout lines derived from the same gestational stage were T-antigen transformed as previously transcribed(Ishov et al., 1999; Ishov et al., 2002). The original ES and ES DaxxΔ260 cells came from J. Michaelson(Michaelson et al., 1999). HEp-2 carcinoma cells and all MEFs were maintained in DMEM supplemented with 10% fetal calf serum (FCS) and 1% antibiotics at 37°C in a humidified 5%CO₂ atmosphere. For experimental use, cells were grown overnight in six-well plates or 35-mm Petri dishes. For immunohistochemical staining, cells were grown on round coverslips in 24-well plates (Corning Glass, Corning, NY)until ∼80% confluence before fixation. For heat-shock treatment, wells or plates were sealed with parafilm and placed in a water bath at 42°C. Cells were either harvested or left to recover at 37°C for various times. To induce stress with heavy metals, cells were incubated in 10 μM, 30 μM,40 μM, 50 μM or 80 μM CdCl₂ for various times. Inhibitors SB203580 and PD98059 (Calbiochem, San Diego, CA) were used at 10 μM and 50μM, respectively. His-tagged SENP-1 plasmids were transfected into HEp-2 cells (Gong et al., 2000)using Superfect (Qiagen, Valencia, CA). Transfected cells were assayed by using an anti-RGS-His antibody (Qiagen, Valencia, CA) and co-stained with different ND10 antibodies.

Cells were lysed either directly in 6 M guanidine-HCl or in a 3:1 dilution of RIPA (150 mM NaCl, 50 mM Tris-HCl (pH 8), 1% NP-40, 0.5% DOC, 0.1% SDS) and SDS sample buffer (250 mM Tris-HCl (pH 6.8), 50% glycerol, 20% SDS, 15%β-mercaptoethanol). Samples were boiled for 5 minutes and proteins were separated by SDS-PAGE, transferred electrophoretically to a nitrocellulose membrane and incubated with rabbit anti-PML antibody (1:2000), rabbit anti-Sp100 antibody (1:2000), monoclonal antibody (mAb) 5.14 anti-human Daxx(Sotnikov et al., 2001)(1:10), rabbit anti-Daxx antibody (Ishov et al., 1999), rabbit anti-mouse-Hsp25 antibody(Tanguay et al., 1993)(1:2500), rabbit anti-Hsp70 antibody (SPA-812) (1:2000) (StressGen Biotechnologies Group, Victoria, Canada) and rabbit anti-heat-shock-factor-1(HSF1) antibody (SPA-901) (1:1000) (StressGen Biotechnologies Group, Victoria,Canada) in 5% milk in PBS/1% Tween as primary antibody. Horseradish-peroxidase-conjugated secondary antibodies (Vector Laboratories,Burlingame, CA) were used to develop the blots. For normalization of loading,all blots were striped and reprobed with anti-tubulin MAb DM 1A (1:20,000 dilution) from Sigma (St Louis, MI).

Cells were fixed in 1% paraformaldehyde(Ascoli and Maul, 1991) for 15 minutes and permeabilized in 0.3% Triton-X. ND10 were visualized using MAb 1150 for Sp100 and MAb 5E10 for PML(Stuurman et al., 1992). Polyclonal rabbit antisera were prepared against a glutathione-S-transferase (GST)-tagged peptide containing PML residues 334-480 (Sotnikov et al.,2001). Human specific anti-Daxx MAb 5.14 were raised against a His-tagged human Daxx fragment from residues 423-740; the binding epitope has been localized to residues 495-535(Sotnikov et al., 2001). HP-1 MAb was obtained from F. Rauscher (The Wistar Institute) and rabbit anti-Daxx antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Human autoimmune serum 21 for fibrillarin(Yasuda and Maul, 1990), human autoimmune serum 44 for lamin B (Maul et al., 1987) and autoimmune serum 822 for coilin were used to test the effects of stress on nuclear integrity. Hsp25, the inducible Hsp70 and HSF1 were analyzed with rabbit anti-mouse Hsp25(Tanguay et al., 1993), rabbit anti-Hsp70 (SPA-812) and rabbit anti-heat-shock factor 1 (HSF1) (SPA-901)obtained from StressGen Biotechnologies Group (Victoria, Canada). Anti-SUMO-1 antibodies came from two sources. The antibodies from M. Matunis(Matunis et al., 1996) had a strong tendency also to stain the nuclear envelope, whereas those produced by Tanguay's group showed strong staining of ND10 and low dispersed nucleoplasmic staining but no significant staining of the nuclear envelope. FITC- or Texas-Red-labeled secondary antibodies (Vector Laboratories) served as fluorescent markers. Cells were examined using a Leica TCS SPII confocal laser-scanning system. Two or three channels were recorded simultaneously and/or sequentially, and controlled for possible breakthrough between the FITC and Texas Red signals and between the blue and red channels. For quantitation of ND10 and ND10-associated proteins, immunohistochemical samples were double-labeled for PML and Daxx; for each sample, the total number of PML dots and the number of colocalizing Daxx dots were counted in 200 cells and transformed using Microsoft® Excel® for graphical representation. Separate Daxx antibody-positive dots, which appear in very few cells and are usually irregularly shaped centromere-associated Daxx deposits (cloud-like domains found in a subpopulation of the cells during recovery), were not included.

Total RNA was extracted from cells using the TRI Reagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer's protocol. The quality of extracted RNA was determined by spectrophotometry and the appearance of characteristic 28S and 18S rRNA fragments on a 1% agarose gel. DNA synthesis was carried out using 5 μg RNA on You-Prime-First-Strand beads (Amersham Pharmacia Biotech, Piscataway, NJ), and oligo dT₁₅(0.5 μg, Promega, Medison, WI), according to the manufacturer's protocols. PCR was performed using 2 μl of the 35 μl cDNA sample. Initial PCR reactions were performed simultaneously for several cycles to determine the range of linearity for each gene. Below, we indicate the genes tested, the size of the expected PCR product, the number of cycles and the sequences of the 5′ and 3′ primers.

• mHsp25 (408 bp; 33 cycles), CTCTTCGATCAAGCTTTCGG,CTCAGGGGATAGGGAAGAGG

• Hsp70 (534 bp; 33 cycles), AAACTCCCTCCCTGGTCTGA,CTTGTCTTCGCTTGTCTCTG

• GAPDH (453 bp; 20 cycles), ACCACAGTCCATGCCATCAC,TCCACCACCCTGTTGCTGTA

All PCR reaction were performed with 1.5 mM MgCl₂, denatured at 94°C for 1 minute, annealed at 63°C for 40 seconds and extended at 72°C for 40 seconds. PCR products were resolved on 2% agarose gel and photographed with a Polaroid camera.

To study the effect of hyperthermic stress on the morphology and composition of ND10, cells were exposed to 42°C for different times. Different nuclear domains such as the nucleolus, coiled bodies and speckles,as well as components of ND10, were visualized by immunohistochemistry using antibodies against PML, SUMO-1, Daxx and Sp100. The distribution of these components was analyzed by confocal microscopy. Triple-labeled cells were used to exclude the possibility that observed changes in ND10 protein distribution were due to a general nuclear breakdown and to exclude potential cell-to-cell variation or effects of microenvironmental differences. In non-heat-shocked control cells, speckles (SC35), Cajal bodies (coilin) and ND10 as labeled by antibodies against Daxx (Fig. 1A) were distributed as previously described, including the frequent association of Cajal bodies with ND10(Ishov and Maul, 1996). In these HEp-2 cells, Cajal bodies and speckles were found at a frequency of 1-7 and 10-20 per nucleus, respectively. Cells tested for the distribution of fibrillarin as a nucleolar label, heat-shock factor 1 (HSF1) and Daxx show the large nucleolar domain labeled, a low nuclear level of dispersed HSF1 as well as the normal Daxx distribution (Fig. 1B). PML and Daxx (Fig. 1C), Sp100 and SUMO-1 (Fig. 1D) co-localized to ND10, except that some large ND10 lacking substantial SUMO-1 accumulation were also observed(Fig. 1D, arrow). The average frequency of ND10 was about six per nucleus in these control cells. After a short (12 minutes) heat shock of 42°C neither speckles nor the nucleolar fibrillarin showed recognizable changes, but Daxx became less prominent in many cells (Fig. 1E) and was absent from ND10-like structures in most other cells(Fig. 1F). HSF1 did not aggregate in cells in which Daxx was released from ND10 after this short interval of thermal exposure (Fig. 1F). The number of PML aggregates increased to 20-25 per nucleus,and Sp100 was present in these additional PML aggregates(Fig. 1G), apparently redistributing with PML. SUMO-1 also disappeared from ND10 in most cells(Fig. 1H). Average domain size was somewhat smaller in heat-shocked cells than in controls, although the larger PML domains were approximately the same size as in controls.

Both SUMO-1 and Daxx staining disappeared from the larger PML aggregates with increasing treatment time, and the staining did not reappear in the smaller ones (shown for 1 hour heat shocks for Daxx in Fig. 1I-K and for SUMO in Fig. 1L). With these longer heat shocks, we did not recognize an obvious change in speckles or in Cajal bodies, but fibrillarin became dispersed by 30 minutes, similar to previous observations (Liu et al.,1996). At this treatment time, we also observed HSF1 aggregation but did not find Daxx localized at these sites(Fig. 1J). The immunohistochemistry-based observations provide evidence that ND10-associated proteins are released from their site of highest concentration, whereas other nuclear compartments (such as speckles and Cajal bodies) remain intact. The release of SUMO-1 and Daxx was relatively rapid (less than 12 minutes in most cells) and occurred in the presence of cycloheximide applied 1 hour before heat shock and therefore did not depend on new protein synthesis (Maul, 1995). These observations suggest that the formation of new PML aggregates is independent of de novo PML synthesis and of the presence of SUMO-1-modified PML. The latter observation is consistent with previous reports that unsumofied PML can bind to nuclear deposition sites(Ishov et al., 1999; Lallemand-Breitenbach et al.,2001).

After 1 hour exposure to 42°C, the release of SUMO-1, Daxx and Sp100 from ND10 appeared to be complete in most cells. Sp100 had totally dispersed in ∼45% of the cells. In the remaining cells, the signal at ND10 had diminished and was detected as many small dots (shown for a SUMO-1- and Sp100-stained cell pair in Fig. 1L). Thus, Sp100 disperses more slowly than SUMO-1 and Daxx. After the initial redistribution of PML into additional and smaller aggregates, the localization of PML did not change substantially during extended periods of elevated temperature exposure. The redistribution of ND10 proteins tested still occurred and Daxx was released in cells pretreated with MG132, an inhibitor of the proteasomal degradation pathway (data not shown), indicating that the disappearance of components out of ND10 was not due to a rapid degradation of proteins.

Cd²⁺ exposure can also induce a stress-like response and dispersion of ND10-associated proteins(Maul et al., 1995). We therefore tested for potential changes in various nuclear domains using protocols similar to those for thermal stress. Unlike the rapid dispersion of SUMO-1 and Daxx and eventually Sp100 in response to hyperthermic exposure, the response to 50 μM Cd²⁺ required ∼2 hours for reliable observation of dispersion in many cells. After 4 hours of Cd²⁺exposure, no changes in speckles, Cajal bodies, nucleoli or nuclear envelope were observed (Fig. 2A,B). HSF1 granules did form and were not associated with Daxx(Fig. 2B). However, Daxx(Fig. 2A,B) and SUMO-1 (not shown) were dispersed. Importantly, PML was also dispersed in most, but not all, cells (shown for Daxx and PML stained cells in Fig. 2C). The retained PML aggregates were smaller and nucleoplasmic staining was increased. In many cells, low numbers of Sp100 aggregates that lacked SUMO-1 or Daxx remained(shown in Fig. 2D for Sp100 and SUMO-1). The apparent simultaneous loss of PML, Daxx and SUMO, and the retention of Sp100 in SUMO-1-negative domains suggest that a mechanism different from that in hyperthermic stress is operating in Cd²⁺-exposed cells. Because PML is the protein that retains Daxx at ND10 (Ishov et al., 1999), PML dispersal might cause the dispersal of other ND10-associated proteins but not necessarily Sp100.

To determine whether the heat-shock-induced dispersion of ND10 is reversible, cells exposed to 42°C for 12 minutes, 1 hour or 3 hours were left to recover at 37°C for 45 minutes, 2 hours, 5 hours and overnight. In all recovery experiments, cells showed reformation of normal ND10 (i.e. recruitment of SUMO-1, Daxx and Sp100 into ND10). However, there was extensive heterogeneity among and within cells. SUMO-1 was not present in all PML-positive sites (shown in the upper left cell of Fig. 1M for 1 hour heat shock and 2 hours recovery, and the cell at the left, in which almost no SUMO-1 staining is recognizable in PML aggregates). Daxx might be present together with some PML aggregates (Fig. 1N, right cell) or not at all(Fig. 1N, left cell). Daxx might also be deposited at heterochromatin without PML like in PML^-/- cells (irregular green deposits right cell, Fig. 1N). After 5 hours of recovery, colocalization was almost completely restored(Fig. 1O). Overall, the number of PML aggregates decreased over time and the frequency distribution of PML aggregates narrowed (Fig. 3A),suggesting release from alternate binding sites that were initially induced by heat shock. By contrast, Daxx aggregates were totally absent after heat shock but increased and colocalized with PML during the study period, although they did not reach the frequency of PML aggregates in the same time(Fig. 3B and directly seen in Fig. 1O for 5 hours recovery,in which some PML aggregates have no Daxx).

In cells exposed to 50 μM CdCl₂ for 6 hours and monitored over 24 hours after medium exchange, ND10 presented their normal shape. After 6-18 hours, the number of mitotic figures was increased. A peak of 17% mitotic cells was recognized at 8 hours recovery and another of 23% at 16 hours recovery, similar in timing if not in magnitude to that observed after heat shock (Maul et al., 1995). Control cells maintained in medium without Cd²⁺ had a mitotic index of 2.1. These findings suggest that CdCl₂-exposed cells were accumulated in a specific phase of the cell cycle and then release during medium exchange, and that the Cd²⁺ concentration used is subtoxic. A higher Cd²⁺ concentration (80 μM) led to cell death during exposures exceeding the length of a cell cycle, with many cells floating off the plate.

Sumolation of PML is important for the recruitment of Daxx into ND10(Ishov et al., 1999). Therefore, the presence of different SUMO-1-modified PML isoforms was analyzed by western blotting. In non-heat-shocked cells, PML antibodies reacted with an isoform migrating at ∼110 kDa and with three major slower-migrating bands(arrows in Fig. 4A),corresponding to three SUMO-1-associated isoforms(Kamitani et al., 1998; Everett et al., 1998; Muller et al., 1998). Upon heat shock, these SUMO-1-associated isoforms gradually diminished, beginning as early as 12 minutes post-exposure. Reprobing the same blot with anti-Sp100 antibodies revealed a selective lost of the major sumofied isoform of Sp100(arrow in Fig. 4C). Finally,use of anti-Daxx mAbs revealed only one major band(Fig. 4B), with a slight decrease in intensity after prolonged heat shock. The slow decay of Daxx is surprising because this protein contains three PEST domains, which are thought to imply fast turnover through proteasome-dependent hydrolysis(Rechsteiner and Rogers,1996). Thus, Daxx might be released from ND10 and bound to other sites, preventing its hydrolysis.

To determine whether the disruption of ND10 upon heat shock is regulated by the activation of a stress-activated kinase pathways, cells were pretreated for 1 hour with SB203580 (Cuenda et al.,1995), which specifically inhibits the stress-activated p38 MAP kinase, or with PD98059, which inhibits the mitogen-activated MEK/ERK pathway(Dudley et al., 1995). Cells were then heat shocked for 1 hour at 42°C and analyzed for the release of PML, Daxx and Sp100. No difference was observed between inhibitor-treated cells and untreated cells. Thus, the rapid heat-shock-induced ND10 modifications in HEp-2 cells appear to be independent of the activation of the p38 MAP kinase and the MEK/ERK pathways.

Similar analysis of cells preincubated for 1 hour with SB203580 followed by 4 hours of Cd²⁺ exposure revealed intact ND10 (shown in Fig. 2E for Sp100/SUMO). Thus,a block in p38 MAPK phosphorylation prevented dispersion of ND10-associated proteins upon Cd²⁺ exposure but not heat shock. Surprisingly,PD98059 had the same effect (shown for Daxx/PML, Fig. 2F) as SB203580 —ND10 were retained over prolonged periods of Cd²⁺ exposure. The different effect of these inhibitors after HS and Cd²⁺ exposure indicate that the release of ND10-associated proteins is regulated by different pathways.

The heat-shock-induced removal of ND10-associated proteins could not be linked to either of two induced signaling pathways, but the release was associated with rapid desumolation of PML. We therefore tested whether a recently described sentrin/SUMO-1 isopeptidase, SENP-1(Gong et al., 2000) might account for the release of ND10-associated proteins. In HEp-2 cells transfected with a SENP-1-expressing plasmid, SENP-1 was found at ND10 and PML was retained in aggregates of the same size(Fig. 5A-C, upper right cell is transfected). No ND10-like staining for SUMO-1 was detected in the SENP-1-transfected cells, even at very low concentrations of SENP-1. SUMO-1 staining was retained at the nuclear envelope, where it presumably modifies RanGAP, suggesting that the SENP-1 effectively and specifically cleaves SUMO-1 at ND10. Daxx was not detected in ND10 even at low expression levels of SENP-1 (Fig. 5E-H). Sp100 was also reduced in most cells but could be found in a few PML aggregates (Fig. 5I-L). It is therefore released much more slowly than Daxx, similar to the finding with heat shock. These findings are consistent with the release of ND10-associated proteins through desumolation observed after heat shock, except that, under these conditions, no new PML sites were formed in the SENP-1-transfected cells. Thus, SENP-1 activity does not induce redistribution of PML to new sites and PML can remain aggregated in the absence of SUMO-1 over extended periods of time. These results point to the role of a SUMO-1-cleaving enzyme,possibly SENP-1, in the release of ND10-associated proteins upon heat shock and suggest that heat shock activates this enzyme rapidly, although this point can not be proved at present.

The supramolecular changes in the nucleus after heat shock suggest that Daxx is released from ND10 and that some PML is distributed to different sites, although much remains in distinct domains. Daxx and PML, as repressors,should have a downstream effect if their release or redistribution has any physiological relevance. However, the potential repressive effect of the heat-shock-induced release of Daxx and PML cannot be directly assayed because many proteins might be inactivated and various transcriptional and splicing processes are affected by heat shock. Also, proteins other than PML and Daxx are released from ND10, complicating the analysis. We therefore decided to test whether the lack of Daxx or PML might affect the induction of two heat-shock proteins, Hsp70 and Hsp25, using PML and Daxx knockout mouse cell lines and their respective controls. The control mouse cell lines respond essentially the same way as human cells with respect to ND10 dispersion after stress (data not shown). Comparison of Hsp synthesis in normal and Daxx^-/- and PML^-/- cells should provide a measure of Daxx or PML contribution to the heat-shock response as measured by the accumulation of heat-shock proteins (in these cells, they cannot be released from a depot). We also used a cell line derived from ES cells(Michaelson et al., 1999) that expresses a Daxx fragment lacking the N-terminal 260 residues, as shown by western blot in Fig. 6. Because the C-terminus is intact, this protein can bind to PML and be recognized by immunofluorescence at ND10 (A.M.I. et al., unpublished), although it lacks the domain involved in binding to other proteins(Ishov et al., 2002). Cells were collected after 1 hour at 42°C and at hourly intervals of recovery at 37°C. Transferred proteins were probed for Hsp25 and reprobed for Hsp70,HSF1 and tubulin after successive striping of the membranes. As shown in Fig. 7, all cell lines showed no difference in the relative amounts of HSF1 and no variation over the time of recovery from 1 hour at 42°C. We also noticed no significant Hsp70 differences between the respective wild-type control and the Daxx and PML knockout cells, although there were minor differences early in the recovery process between sets of cells, which were derived from different stages in development. The highest accumulation of either Hsp70 or Hsp25 was seen after 6 hours of recovery. However, unlike Hsp70, Hsp25 differed dramatically between the normal and knock out cells. In Daxx^-/- cells, a substantial amount of Hsp25 was already present in cells not exposed to higher temperatures and increased from this high level, whereas, in normal cells,substantial amounts of Hsps were only seen after 2 hours of recovery. A similar pattern was observed in PML^-/- cells. This pattern was different in the cells containing an N-terminally truncated Daxx, in which Hsp25 was barely detectable at 6 hours of recovery and normal cells produced more Hsp25. The results suggest that, in the absence of Daxx or PML, Hsp25 is produced in appreciable amounts, whereas, in the presence of a truncated form of Daxx, the production of Hsp25, but not Hsp70, is inhibited. This suggests that Daxx, presumably through its release-based availability, might bind to some other protein to inhibit Hsp25 production, whereas the truncated Daxx retains its repressive activity, which might even be enhanced.

The inhibition of Hsp25 production might happen at the transcriptional or translational level. To test whether Daxx or PML have an influence on the transcription of Hsp25, cells exposed for 1 hour to 42°C were analyzed for accumulation of Hsp25 mRNA over a range of recovery times using semi-quantitative RT-PCR. For normal and PML^-/- cells, induction of Hsp70 mRNA was the same, but Hsp25 mRNA increased slowly over time in the PML^+/+ cell. Hsp25 mRNA was present even in unstimulated PML^-/- cells and levels increased after stimulation and recovery (Fig. 8A). Again,in Daxx^+/+ cells and Daxx^-/- cells, there was no significant difference in the Hsp70 mRNA concentration or accumulation (Fig. 8B) but Hsp25 mRNA accumulation in Daxx^+/+ cells increased gradually from an insignificant amount before hypothermic exposure to high concentrations after 4 hours of recovery, whereas Hsp25 mRNA in Daxx^-/- cells was at maximum after only1 hour exposure and before any recovery time. Message, although at low levels, was also detectable before any exposure to elevated temperatures. Thus, the presence of Daxx and PML, and potentially their heat-shock-induced distribution, might inhibit the transcription and consequent translation of Hsp25. Importantly, the increased transcription during recovery correlates with the resegregation of Daxx and PML into ND10 during recovery and thus with the unavailability of Daxx and PML.

CdCl₂ exposure induced the stress response and dispersed PML and, along with it, most of the other ND10-associated proteins in most cells. We therefore tested what effect lack of PML would have on the CdCl₂-induced heat-shock response. As shown in Fig. 9, exposing cells to various concentrations of CdCl₂ for 6 hours decreased slower-migrating PML isoforms or sumolated species, resulting in a higher concentration of the fastest-migrating species. These should become available for binding in the nucleoplasmic space. The exposure of cells to CdCl₂ had no effect on the amount of cyclin D1 and tubulin or of HSF1, which might have influenced transcription of Hsp genes. However, the apparent amount of Hsp70 was higher than that of Hsp25 in PML^+/+cells; that is, the ratio of Hsp70 to Hsp25 produced in CdCl₂-treated cells was reversed when compared with the heat-shock induction of these proteins. In CdCl₂-exposed PML^-/-cells, the same ratio of induction was found as with heat shock (i.e. less Hsp70 than Hsp25). Proteins were separated on the same gel and the transferred proteins probed on the same membrane after successive stripping to ensure that the analysis shows comparable concentrations. These results indicate that hyperthermia and CdCl₂ in the absence of PML have different effects on Hsp induction.

Cellular products stored in organelles, such as lipid droplets or glycogen granules, are not functional at these sites. Aggregation and release of these products are triggered by demand for them, often through signal-transduced secondary modification of various enzymes. The nucleus contains sites (ND10)where certain proteins are aggregated and maintained through the interacting functions of PML (Ishov et al.,1999; Lallemand-Breitenbach et al., 2001; Zhong et al.,2000a). We suggested that these organelles might function as depots, or regulatable storage sites(Negorev and Maul, 2001). At least one function of a depot (recruitment) was shown to be regulated by transcriptional activation (interferon) and enzymatic modification(sumolation) (Chelbi-Alix et al.,1995; Grotzinger et al.,1996; Ishov et al.,1999; Stadler et al.,1995). Here, we have identified two possible mechanisms induced by environmental stress that result in another depot function (release) by demonstrating that this function is also specifically regulated and influences the stress response. Our data show that stresses in the form of moderate heat shock or exposure to nontoxic CdCl₂ concentrations change the distribution of ND10-associated proteins but do not overtly damage the nucleus. Only the fibrillarin distribution in the nucleolus changes after about 30 minutes of thermal stress at 42°C(Liu et al., 1996). Also the heat-shock factor HSF1 changes its distribution from a diffuse to a more aggregated form. All of these changes are reversible and cells resume cycling in a somewhat synchronized fashion.

Environmental insults such as hyperthermia, ultraviolet radiation, chemical shock and heavy-metal exposure are transduced from the cell surface to the nucleus of the cell through stress-activated signaling pathways. However,heat-shock-induced dispersion of ND10 proteins, such as SUMO-1, Daxx and Sp100 was not inhibited by blocking the ERK1/2 and p38 MAPK pathways, making it unlikely that the signaling pathways inducing Hsp synthesis are involved in ND10 protein dispersion after heat shock. Instead, a SUMO-1 isopeptidase, such as SENP-1, appears to be rapidly activated, because no protein synthesis is necessary, and SUMO-1, Daxx and Sp100 but not PML were released. A potential cellular defense response located at ND10 might be faster than or precede that of the Hsp synthesis.

Unlike heat shock, Cd²⁺ exposure induced the dispersion of PML but not Sp100, suggesting the involvement of a mechanism to disperse PML. The block in Cd²⁺-induced ND10 dispersion by inhibiting either p38 MAPK or ERK1/2 implicates both stress and mitogen signaling pathways in the dispersion. The apparent communication between the stress-and mitogen-induced signaling pathways for the release of proteins from ND10 by Cd²⁺exposure might resemble that reported for arsenite, in which ERK activation is strongly enhanced through p38 MAPK (Ludwig et al., 1998; Rouse et al.,1994). Alternatively, both pathways might require the same substrate, as shown for Mnk1/2 after TPA(12-O-tetradecanoylphorbol-13-acetate) treatment(Waskiewicz et al., 1997),which would provide a more complex or targeted signaling and response to Cd²⁺ exposure. The most direct route for such a possibility would involve phosphorylation of PML at the three consensus sequences for MAP kinases (ψX(S/T)P), where ψ is an aliphatic amino acid(Alvarez et al., 1991), which are conserved between human and mouse PML.

The effect of Cd²⁺ and heat-shock exposure on the release of ND10 proteins resembles that induced by viruses during lytic infection. In fact, both Cd²⁺ and heat-shock exposure reactivate herpes simplex virus from latency. However, the two stress inducers appear to differ in the mechanisms that mediate the release. Activation of a specific SUMO isopeptidase such as or similar to SENP-1 appears to be an immediate event in the heat-shock response, whereas downstream effects of the mitogen and stress signaling pathways induced by Cd²⁺ occur later. Both effects on ND10 are apparently independent of the classical heat-shock response, because neither pretreatment by elevated temperatures nor use of protein synthesis inhibitors influences their actions (Maul et al., 1995). Our results suggest that both stressors release ND10-associated proteins by acting on the structure of ND10 at different levels. The MEK-dependent dissociation of PML by Cd²⁺ might cause the removal of proteins associated with PML. This PML dispersal might occur through direct phosphorylation of PML. Heat shock, however, does not interfere with PML as a scaffold protein of ND10 but by changing the sumolation level of PML and Sp100, thereby reducing the ability of PML to retain other components. Thus, herpes simplex virus reactivation from latency by the two different stresses might be a consequence of specific proteins released from ND10 and not of a single, specific, stress-related mechanism.

The complexity of the stress response makes it difficult to analyze whether a directly observed effect such as release of specific proteins from a nuclear domain has physiologically relevant consequences. More than one protein is released into a potentially available pool when such multiprotein aggregates are dispersed. Adding more proteins, such as Daxx or PML, through transient expression would also increase their individual availability but it also induces further aggregation and titration of a host of other effector proteins as well as the stress of the transfection procedure. The availability of PML^-/- and Daxx^-/- cells and their direct wild-type counterparts from equivalent gestation stages of the mouse enabled analysis of the effect of lack of PML or Daxx on the induction of the Hsp synthesis. Surprisingly, different effects were found for Hsp25 and Hsp70 after hyperthermic treatment of cells. After heat shock, Hsp70 synthesis appeared to be independent of the presence or absence of PML and Daxx, suggesting no general involvement of the heat-shock factor with PML or Daxx. However, Hsp25 synthesis was suppressed by the presence of PML and Daxx in wild-type cells. This suppression was even more evident in cells expressing a truncated Daxx that can bind to PML. The suppressive effect is presumably exerted at the transcriptional level because accumulation of Hsp25 mRNA was very slow in the presence of PML and Daxx but immediate and near maximal in the absence of PML or Daxx. The removal of Daxx and PML from the freely accessible compartment into the bound and therefore inaccessible compartment of ND10 during the recovery period appears to have the same effect as the absence of these proteins. It correlates with the full expression of HSP25. This is consistent with the idea that release of Daxx induces repression of Hsp25. In cells exposed to CdCl₂, we find the same pattern for Hsp25 as for thermal stress. It remains unclear how the release of either PML or Daxx can have the same effect of reducing or slowing transcript accumulation. Daxx might still affect the production of Hsp25 in the absence of PML by being released from its alternate binding sites. However, Hsp70 production was strongly inhibited in the absence of PML during CdCl₂ exposure in PML^-/-cells. PML might have to be released, suggesting that PML can modulate the activation of Hsp70 production. We suggest that the regulated release of Daxx and PML by various types of stress modifies the cell's heat-shock response differently at the transcriptional level. It is astonishing that the cell has mechanisms to slow down the heat-shock response.

This study was supported by funds from NIH AI 41136 and NIH GM 57599, the Human Frontier Science Program, the G. Harold and Leila Y. Mathers Charitable Foundation and the Canadian Institutes of Health Research MOP-43958 (R.M.T.). We thank P. Pandolfi for PML^-/- cells and S. Lowe for antibodies. NIH Core grant CA-10815 is acknowledged for support of the microscopy and sequencing facility.