Interactions between extracellular signal-regulated protein kinase 1, 14-3-3epsilon, and heat shock factor 1 during stress.

Cytoprotection during the heat shock response is a complex phenomenon involving multiple inducible mechanisms. We have examined the interaction of two key molecular components in the response, heat shock transcription factor 1 (HSF1) and extracellular signal regulated protein kinase (ERK). Whereas both HSF1 and ERK are required to protect cells against apoptosis, ERK activation is paradoxically antagonistic to trans-activation of hsp promoters by HSF1 and HSP accumulation during heat shock. We have found that the two pathways interact directly and that heat shock causes the physical association of ERK1 with HSF1, an interaction that promotes the kinase activity of ERK in heat-shocked cells. ERK activation results in the recruitment of the phosphoserine binding protein 14-3-3epsilon in a manner dependent on previous HSF1 phosphorylation by ERK. The effects of 14-3-3epsilon binding on HSF1 were complex, however, depending on extracellular conditions, in that HSF1-14-3-3 binding at 37 degrees C led to the cytoplasmic sequestration and repression of HSF1, whereas heat shock overrode these effects and caused quantitative nuclear localization of HSF1. Although the effects of 14-3-3epsilon binding to HSF1 were overridden acutely by stress, during recovery from heat shock, 14-3-3epsilon association again led to enhanced cytoplasmic localization of HSF1, implicating a role for ERK/14-3-3epsilon in HSF1 deactivation in recovering cells. Association of HSF1 with ERK and 14-3-3epsilon during heat shock may thus modulate the amplitude of the response and lead to efficient termination of HSP expression on resumption of growth conditions.

The heat shock response is induced by stress and protects cells from a range of extracellular stresses (1)(2)(3). Such cytoprotection is complex and involves a number of factors, most notably the induction of heat shock protein molecular chaperones (HSP) 1 and the activation of extracellular signal-regulated pro-tein kinase (ERK) activity (1)(2)(3)(4). It is odd that although both HSP induction and ERK activation result in cytoprotection, ERK is known to be a repressor of HSP synthesis (5,6). Heat shock transcription factor 1 (HSF1) regulates HSP synthesis and activates hsp gene transcription (7)(8)(9). Inactivation of the murine hsf1 gene results in a complex phenotype indicating an essential function for hsf1 in growth, development, and shortterm response to stress (10). Disruption of hsf1 (hsf1 Ϫ/Ϫ ) in mouse embryonic fibroblasts leads to a profound loss of thermotolerance and markedly increased susceptibility to heatinduced apoptosis (10,11). Under normal conditions, cellular HSF1 exists in a transcriptionally repressed state (12,13). Such HSF1 is monomeric, constitutively phosphorylated, and unable to bind the cis-acting heat shock elements located in the promoters of hsp genes (14,15). Induction of transcriptional activity by heat shock then results in the conversion of HSF1 from inactive monomer to a DNA-binding trimer (12,16,17). Activation of HSF1 by stress is a multistep process involving trimerization, acquisition of heat shock element-binding activity, novel phosphorylation, and trans-activation of hsp genes (14,15,18). Trimerization of HSF1 is governed by leucine zipper domains in the N terminus and is subject to intramolecular negative regulation by a fourth leucine zipper domain in the C terminus (12). The molecular chaperone HSP90 functions as the principal repressor of HSF1 in unstressed cells and plays a major role in retaining HSF1 in an inactive state; HSF1 trimerization is accompanied by the sequestration of HSP90 in protein aggregates and escape from HSP90-containing HSF1 complexes in response to stress (13).
The ERK pathway is regulated downstream of a number of transmembrane receptors and couples extracellular signals to intracellular events controlling the activity of transcription factors that mediate cell proliferation (19 -21). ERK itself is a component of a three-kinase cascade at the membrane consisting of the sequential activation of a mitogen-activated protein kinase kinase kinase (Raf-1), a mitogen-activated protein kinase kinase (MEK2) and finally ERK/MAPK itself (19,20). Phosphorylation by MEK2 is an essential step in ERK activation (20). Raf-1 is activated at the membrane by association with Ras proteins (22). Ras is in turn coupled to transmembrane receptor occupancy through receptor tyrosine autophosphorylation and recruitment of phosphotyrosine-binding adaptor proteins that activate Ras and permit it to associate with Raf-1 at the membrane (23). As mentioned below, ERK activation downstream of growth factor binding represses HSF1 (24). HSF1 seems to exert a reciprocal influence on cell proliferation in that HSP70, a major product of HSF1 transcriptional activ-ity, inhibits the ERK pathway through effects on at least one step in the pathway, the Raf-1 step in ERK signaling (25).
HSF1 is also subject to additional layers of regulation, including up-regulation through binding the apoptosis modulator DAXX and modulation by ERK (26 -28). Our studies, as well as the work of others, have shown the hierarchical phosphorylation of human HSF1 within its transcriptional regulatory domain by ERK1 (on serine 307) and by GSK3 (on serine 303) (6, 29 -32). The regulatory domain of HSF1 functions as a molecular switch coupling hsp gene transcription to cellular conditions, repressing C-terminal trans-activation domains under growth conditions, and causing powerful stimulation of the same activation domains under stress conditions (30,33,34). ERK1 and GSK3 phosphorylate HSF1 on serine residues within a proline-rich region (RVKEEPPS 303 PPQS 307 PRV) of the regulatory domain (6,29). Recent studies show that phosphorylation on serine and threonine is converted into an intracellular signal by association with regulatory proteins that recognize serine/threonine phosphorylated domains (35). The first such proteins to be identified were those of the 14-3-3 family, which bind to a wide array of cellular proteins (36,37). At least seven 14-3-3 genes exist in vertebrates, and these give rise to nine protein isoforms (␣, ␤, ␦, ⑀, ␥, , , , ) (38). 14-3-3 proteins are predominantly dimeric within the cell and bind either to multiple sites within single proteins such as c-Raf1 or act as a bridge, with one 14-3-3 dimer binding to two different proteins (39). 14-3-3 dimers may thus act as molecular scaffold proteins, bringing together proteins that interact functionally and effecting phosphorylation-dependent cell regulation (39). We have shown that HSF1 binds to 14-3-3 after exposure of G o cells to mitogens (24). ERK activation by mitogenic stimulation and consequent phosphorylation of serines 303 and 307 led to direct HSF1 association with 14-3-3⑀ and 14-3-3. 14-3-3 binding inhibits both the transcriptional activity and nuclear accumulation of HSF1 in HeLa cells at 37°C (24). These experiments showed that 14-3-3⑀ mediates interaction between phospho-Ser-307, Ser-303-HSF1, and CRM1 and thus contributes to cytoplasmic sequestration and transcriptional repression after ERK activation.
In this study, we have examined the impact of heat shock on this form of cell regulation. We find that the association of both ERK and 14-3-3 (⑀ and ) with HSF1 is strongly activated by heat shock. Binding of each protein is largely independent of the other and high avidity binding of each protein requires serines 303 and 307 of HSF1. HSF1-ERK association in heatshocked cells promotes ERK activity and leads to phosphorylation of HSF1 itself on serine 307. This phosphorylation in turn promotes HSF1 association with 14-3-3⑀. Our data do not support the existence of a ternary complex between HSF1, ERK, and 14-3-3⑀ and instead suggest that the proteins may compete for binding to the region of HSF1 spanning serines 303 and 307. HSF1 association with 14-3-3⑀ in heat-shocked cells seems to be involved in the attenuation of HSF1 activity during recovery and leads to accelerated cytoplasmic localization of HSF1.
Cell Culture and Transient Transfection-HeLa and SCC7 cells were cultured in Ham's F-12 medium (Mediatech, Inc.) supplemented with 10% heat-inactivated fetal bovine serum. NIH-3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-in-activated FBS. HeLa cells (2.5 ϫ 10 5 cells/well) in six-well plates were transfected with the plasmids indicated in the figure legends in triplicate using FuGENE6 (Roche) according to the manufacturer's protocol as described previously (24,42). pCMV-lacZ plasmid was co-transfected as an internal control for transfection efficiency. pcDNA3.1 empty vector was used as a blank plasmid to balance the amount of DNA introduced in transient transfection. Cells were harvested after 24 h of transfection. Luciferase and ␤-galactosidase activity assays were then performed according to the Promega protocol. Luciferase activity was normalized to ␤-galactosidase activity. Results were expressed as relative luciferase activity of the appropriate control.
Mitogen-activated Protein Kinase and S6 Kinase Assays-After experiments, cells (2 ϫ 10 6 ) were washed three times in ice-cold phosphate-buffered saline, pH 7.4, and lysed in 1.0 ml of Buffer A (10 mM KH 2 PO 4 , 1 mM EDTA, 5 mM EGTA, 10 mM MgCl 2 , 50 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 2 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin, and 4 mM p-nitrophenylphosphate, pH 7.0) with the use of a Dounce homogenizer. Samples were clarified at 60,000 rpm, and supernatants were aliquoted for subsequent protein and kinase assays before snap-freezing in liquid nitrogen. For ERK/ MAPK assay, 5 l of sample was added to 25 l of MAPK Buffer B (60 mM ␤-glycerophosphate, 30 mM p-nitrophenylphosphate, 25 mM MOPS, 15 mM MgCl 2 , 150 M [ 32 P]ATP, 0.1 mM sodium orthovanadate, and 5 mM EGTA, pH 7.2) containing either myelin basic protein at 200 g/ml or consensus MAPK peptide (APRTPGGRR; Upstate Biotechnology, Lake Placid, NY) at 1 mM. Incubations were performed for 20 min at 30°C. Samples were then collected on P81 filters; filters were allowed to dry and were then washed three times in 10 ml of ice-cold phosphoric acid. Filters were counted by Cerenkov counting. Assays were carried out on triplicate samples. For S6 kinase assay, samples were prepared as above and assayed in Buffer B plus 250 M protein kinase inhibitor (PKI), 1 mM dithiothreitol, and 250 M consensus S6 kinase peptide (RRRLSSLRA; Upstate Biotechnology). Assays were carried out as for ERK.
In Vitro Translation and GST Pull-down Assay-GST fusion protein, 14-3-3⑀-GST, and control GST protein were purified from bacterial lysates using standard methods. HSF1wt and mutant were produced by in vitro translation in the presence of [ 35 S]methionine (Ͼ1000 Ci/ mmol; PerkinElmer Life and Analytical Sciences) according to manufacture's protocol using the T7-TNT quick-coupled transcription/translation system (Promega) with pcDNA3.1 HSF1wt and pcDNA3.1 HSF1 303A/307A expression plasmid. GST alone was used as a negative control. 14-3-3⑀-GST or GST (total protein, 1.5 mg) were coupled to glutathione-Sepharose beads (Amersham Biosciences) and washed with the appropriate extract buffer containing protease inhibitors. The beads were then incubated with various in vitro-translated 35 S-labeled proteins at 4°C for 1 h. Finally, the beads were washed extensively and analyzed by 10% SDS-PAGE and x-ray film autoradiography.
Immunofluorescence Microscopy-Immunofluorescence microscopy and co-localization studies were carried out as described previously (24). In brief, after transfection of HeLa cells growing on four-chamber tissue culture slides with GFP-HSF1 constructs or HA-14-3-3⑀, cells were fixed with 4% paraformaldehyde in phosphate-buffered saline. Fixed cells were incubated with primary antibodies, anti-HA, anti-14-3-3⑀, and antibody binding was detected using Texas Red-conjugated secondary antibody (Amersham Biosciences). In all experiments, cells were then counter-stained with 4,6-diamidino-2-phenylindole (Roche) to visualize the nucleus. Images were acquired using a Nikon Eclipse E600 microscope equipped with a RT color SPOT digital camera and processed using SPOT software (Diagnostic Instruments, Inc.).

RESULTS
ERK Activity in Heat-shocked Cells-We first examined the kinetics of ERK activation by heat in murine and human cells (Fig. 1). We measured the electrophoretic mobility of ERK1 and ERK2 from 0 to 60 min of heat shock at 43°C as an indicator of upstream activation of the kinase (Fig. 1A). Activation of ERK1 and ERK2 is accompanied by phosphorylation through the upstream kinase MEK2 and results in decreased electrophoretic mobility (20). There was a rapid (within 10 min) and prolonged electrophoretic mobility shift in ERK1 and ERK2 in NIH-3T3, HeLa, and SCC cells that persisted until 60 min of heating (Fig. 1A, lanes 3-5, anti-ERK1/2 immunoblot). The electrophoretic mobility shift in ERK after heat shock was not as profound as the shift induced by exposure of cells to the positive control condition using 100 M phorbol 12-myristate 13-acetate (Fig. 1A, lane 2, anti-ERK1/2 immunoblot). In addition to ERK, we also probed the blots for HSF1, which also underwent a profound shift in electrophoretic mobility detectable after 10 min and increasing to 60 min of heat shock (Fig.  1A, lanes 3-5, anti-HSF1 immunoblot). Phorbol 12-myristate 13-acetate also induced a mobility shift in HSF1, most noticeably in NIH 3T3 cells, but less profound than the effect of heat shock (Fig. 1A, lane 2, anti-HSF1 immunoblot), which also underwent a larger electrophoretic mobility shift after heat shock (Fig. 1A, lanes 3-5, anti-HSF1 immunoblot). Alterations in ERK electrophoretic mobility were correlated with a large increase in ERK kinase activity in extracts from the heatshocked cell (Fig. 1B). Using two substrates, ERK activation was observed in each case by 10 min and slightly preceded the onset of a large increase of p90 S6 kinase (RSK2) activity assayed in parallel in the same extracts (Fig. 1B).
Heat Shock Causes the Association of HSF1 with ERK1 and Activation of Cellular ERK-Because the kinetics of ERK activation and HSF1 hyperphosphorylation are strongly correlated, we examined potential interaction between these two molecules ( Fig. 2). We have examined the binding of HSF1 to an ERK1-GST fusion protein overexpressed in cells, followed by GST pull-down in cell extracts (Fig. 2). Examining the top, we observed minimal ERK-HSF1 association in unstressed cells ( Fig. 2A, lane 1), but detected binding immediately after 60 min at 43°C (Fig. 2A, lane 2) when ERK activity is high (Fig.  1). Binding decays again on recovery at 37°C for 6 h after heat shock ( Fig. 2A, top, lane 3). This pattern is amplified when HSF1 is overexpressed in the cells ( Fig. 2A, lanes 4 -6). We also examined a briefer exposure of the film to luminescent antigen to compare relative HSF1-ERK1 association between cells overexpressing HSF1 ( Fig. 2A, second blot). Under these assay conditions, HSF1-ERK1 association in wild-type cells was not seen ( Fig. 2A, lanes 1-3). However, this exposure illustrates the point that binding is highest in heat-shocked cells overexpressing wild-type HSF1 ( Fig. 2A, HSF1wt, lane 5). We compared binding to HSF1wt with a mutant form of HSF1 with serines 303 and 307 mutagenized to alanine ( Fig. 2A, lanes 7-9); serine 307 is an in vivo substrate for ERK, and the region spanning serines 303 and 307 has been shown to be a binding site for 14-3-3⑀ (24). Mutation of serine 303/307 to alanine inhibited HSF1 association with ERK by 60 -70% ( Fig. 2A, lanes 7-9). We were also able to detect the molecular chaperones HSP70 and HSP90 in the complex ( Fig. 2A). HSP70 was highest in the cells recovering from heat when intracellular HSP70 levels are expected to be highest. HSP90 levels in the ERK1-HSF1 complex were lowest immediately after heat shock when ERK1-HSF1 association is highest ( Fig. 2A). It is not clear from these data whether the HSP90 is directly associated with ERK or indirectly through HSF1. However, HSP90 is a known HSF1-associated protein and dissociates from HSF1 after heat shock (13). We also probed for the presence of 14-3-3 in the complexes as 14-3-3⑀ has been shown to bind to the Ser 303/307 region of HSF1; however, no evidence of 14-3-3 within the GST-ERK protein complex was found by probing immunoblots of the GST-HSF1 pull-down proteins with pan-14-3-3 antibodies or antibodies to 14-3-3⑀ (data not shown). We were able to detect FIG. 1. Effects of heat shock on ERK1, ERK2, and HSF1. A, cells (NIH 3T3, HeLa, and SCC) were either untreated, exposed to 100 M phorbol 12-myristate 13-acetate, or heat-shocked for 10, 20, or 60 min at 43°C (HS). Cultures were then rinsed in ice-cold phosphate-buffered saline, lysed in Laemmli sample buffer, and analyzed by immunoblot with anti ERK or anti-HSF1 antibodies. Experiments were repeated twice. B, we examined the activity of the ERK and RSK2 kinases in extracts from NIH 3T3 cells from 5-60 min of heat shock at 43°C. Extracts were prepared and analyzed as described under "Materials and Methods." We used either myelin basic protein (MBP) or a consensus MAPK peptide to assay ERK kinase activity and RSK2 peptide in the S6K RSK2 assay. Assays were carried out in triplicate, and experiments were performed three times with reproducible results. the increased phosphorylation of ERK1 in the complexes immediately after heat shock using anti-phospho-ERK antibodies ( Fig. 2A, bottom). This indicates that HSF1-ERK1 complexes are likely to have ERK kinase activity.
We next examined activation of ERK in the bulk cytosol of cells with or without HSF1 overexpression, using levels of phospho-ERK1/2 (pp-ERK1/2) as an index of ERK activation (Fig. 2B). Overexpression of HSF1 did not affect pp-ERK1/2 levels in unstressed cells in which pp-ERK1/2 remained minimal. Heat shock caused an increase in pp-ERK1/2 levels (Fig.  2B, lane 1), in accordance with the data in Fig. 1, and this increase was strongly amplified by HSF1 overexpression (Fig.  2B, lane 5). Single mutations of serines 303 and 307 to alanine did not affect the ability of transfected HSF1 to activate ERK phosphorylation, whereas double mutation (S303A/S307A), a modification that inhibits HSF1-ERK binding ( Fig. 2A), blocked the increase in pp-ERK1/2 levels after heat shock (Fig.  2B). The effect of HSF1 on ERK activity was not mimicked by the related transcription factor HSF2, and overexpression of HSF2 did not significantly increase pp-ERK1/2 levels after heat shock (Fig. 2C, lane 3), whereas HSF1, as shown above, was very effective (Fig. 2C, lane 4). Transfection of a chimeric construct with residues 1-327 of HSF1 fused to the C terminus of residues 344 -536 of HSF2 also activated ERK, implicating the amino-terminal region of HSF1, which contains serines 303 and 307 in the stimulation of ERK activity (Fig. 2C, lane 2). Because our previous studies showed that growth factor-stimulated ERK activation leads to the association of 14-3-3 with HSF1 (24), we examined next whether a similar effect occurs after heat shock.
We also carried out the binding experiments in reverse, using HSF1 to immunoprecipitate 14-3-3 (Fig. 3B). Cells were transfected with HSF1-FLAG to aid the recovery of 14-3-3, and then the lysates from controls or heat-shocked cells were analyzed by immunoprecipitation with anti-FLAG or a control IgG (Fig. 3B). When immunoprecipitates were analyzed with anti-HSF1 antibody, we detected HSF1-FLAG in the anti-FLAG immunoprecipitates, and HSF1 was in its hyperphosphorylated form after heat shock (Fig. 3B, I, a). We were able to detect 14-3-3 association with HSF1 in the heat-shocked fraction when immunoprecipitates were probed with a pan-14-3-3 antibody (Fig. 3B, II, b). As can be seen from the whole cell extract (WCE) control blot, the pan-14-3-3 antibody detects 14-3-3 in two main bands (Fig. 3B, II, b) and both bands are seen in association with HSF1 (Fig. 3B, I, b). These species included 14-3-3⑀ and 14-3-3, as indicated by blotting with antibodies specific for these 14-3-3 isoforms (Fig. 3B, c and d).

FIG. 2. Heat shock induces ERK1 association with HSF1. A,
HeLa cells stably expressing ERK1-GST were transfected transiently with either empty vector pcDNA3.1 (Vector), wild-type HSF1 (HSF1wt), or pHSF1mut (serine 303/307-alanine mutation) (HSF1mut). Cultures from each of these three conditions were then lysed in immunoprecipitation buffer without further treatment (lanes 1, 4, and 7), immediately after heat shock at 43°C (HS) (lanes 2, 5, and 8) or after 60 min at 43°C and 6 h recovery at 37°C (HS/R) (lanes 3, 6, and 9). ERK1-GST was recovered from the lysates by GSH-agarose affinity chromatography and analyzed by 10% SDS-PAGE. Eluates were then analyzed for the presence of HSF1, HSP70, and HSP90 by immunoblot using chemiluminescence/x-ray film autoradiography (Materials and Methods). Levels of ERK1-GST and ppERK1-GST were assayed by immunoblot assay of bulk extracts with specific antibodies. HSF1 blots were analyzed at two different exposures with x-ray film. The longer exposure (top blot) shows ERK1-GST association with HSF1 in untransfected cells (lane 2); the briefer exposure (next blot) illustrates the relative activation of  3. Effect of heat shock on association of HSF1 with 14-3-3. A, cells were stably transfected with 14-3-3⑀-GST and then transiently transfected with empty vector pcDNA3.1 (Vector), with pHSF1 (HSF1wt) or with pHSF1mut (serine 303/307-alanine mutation) (HSF1mut). Replicate cultures were then either left at 37°C (Ϫ) or heat-shocked at 43°C for 60 min (HS). Cultures were then rinsed in ice-cold phosphatebuffered saline and lysed in extraction buffer before GST pull down assay as described under "Materials and Methods." 14-3-3⑀-GST recovered from the lysates by GSH-agarose affinity resin was analyzed by 10% SDS-PAGE. A, I, eluates from the 14-3-3⑀-GST were then analyzed for the presence of HSF1 and HSP70 by immunoblot assay using chemiluminescence/x-ray film autoradiography. Whole cell extracts (A, II) were also probed for the presence of HSF1, HSP70, 14-3-3⑀-GST, and endogenous 14-3-3⑀. Experiments were carried out three times with similar results. B, 14-3-3-HSF1 interactions were probed in a different way, by co-immunoprecipitation analysis, in this case using cells transfected with pHSF1-FLAG. Cells were transiently transfected with pHSF1-FLAG and then HSF1-FLAG containing complexes isolated in extraction buffer as described under "Materials and Methods." To control for the specificity of HSF1-FLAG enrichment by anti-FLAG antibody, cell lysates were also probed with a control IgG that does not recognize the FLAG epitope (control IgG). Complexes were then probed by immunoblot analysis for the presence of HSF1-FLAG, total 14-3-3 and 14-3-3 isoforms 14-3-3⑀ and 14-3-3 using specific antibodies to probe the blots (B, I). Control blots (B, II) were also
It is interesting that these properties of heat shock (the enhancement of intracellular HSF1-14-3-3 binding) were also observed under cell-free conditions in vitro using GST pulldown of in vitro-translated HSF1 with recombinant 14-3-3⑀-GST (Fig. 3D, compare lanes 3 and 4). HSF1 binding to 14-3-3⑀-GST was induced by heat shock in the presence of added protein kinases ERK2 and GSK3␣, although binding was also detected in the absence of added kinases (Fig. 3E, lanes 13-16).
Binding of HSF1 to 14-3-3⑀ was specific as judged by failure of HSF1 to bind to the GST control beads (Fig. 3E, lanes 1-8). It is possible that heat shock enhances 14-3-3⑀ binding to the Ser303/307 region of HSF1 as well as exposing a second binding site in HSF1. 14-3-3⑀ could bind to this site in a phosphorylation-independent manner as has been observed for some 14-3-3 targets. We therefore examined the ability of wild-type (wt) HSF1 and mutant (mut) HSF1 with both serines 303 and 307 mutated to alanine to bind GST-14-3-3⑀ (Fig. 3F). Although both forms of HSF1 are efficiently transcribed and translated in vitro (Fig. 3F, lanes 5 and 6), wt HSF1 binds to 14-3-3⑀-GST with markedly greater avidity (Fig. 3F, lanes 3 and 4). However, the HSF1 mutant was observed to bind in repeated experiments, albeit at ϳ20% of the efficiency of wtHSF1. Our experiments indicate that serines 303 and 307 are required for efficient HSF1 binding to 14-3-3⑀. They also suggest the possibility that 14-3-3⑀ can bind with reduced efficiency to the unphosphorylated Ser-303/307 region or to another as-yet uncharacterized site in HSF1 (Fig. 3F, lanes 3 and 4). Similar conclusions can be drawn from the in vivo experiment in (Fig. 2A).
Given that the association of 14-3-3⑀ with HSF1 has been shown to inhibit the activation of the HSP70B promoter in cells responding to serum stimulation (24), we examined whether a similar effect was operative during heat shock. Cells were transfected with the HSP70B promoter construct pGL.HSP70B and then subjected to heat shock either without HA-14-3-3⑀ expression or with expression of increasing quantities of HA-14-3-3⑀ (Fig. 5). HA-14-3-3⑀ transfection, although leading to enhanced HA-14-3-3⑀ expression, did not repress the heatinduced activity of the HSP70B promoter (Fig. 5). This finding is in contrast to our earlier studies on mitogen repression of HSF1 showing that HA-14-3-3⑀ overexpression effectively inhibits this promoter in the absence of stress (24). Clearly heat shock is able to override the inhibitory effects on HSF1 of 14-3-3⑀ association.
Because our previous studies showed that 14-3-3⑀ overexpression represses HSF1 activity and HSF1-DNA binding by cytoplasmic sequestration of HSF1 (24), we examined whether this effect occurs under stress conditions (Fig. 6). Previous studies indicate that HSF1 is normally distributed between the cytoplasm and a diffuse nuclear localization in the absence of stress (24). Serum stimulation or 14-3-3⑀ overexpression drives HSF1 into an almost exclusively cytoplasmic localization, and this effect is blocked by mutagenesis of serines 303 and 307 (24). By contrast, after heat shock, HSF1 was redistributed to an almost exclusively nuclear distribution, with a proportion of carried out on whole cell extracts (WCE) from the transfectants. C, cells were stably transfected with 14-3-3⑀-GST and then transiently transfected with empty vector pcDNA3.1 (Vector), with pHSF1/HSF2 chimera (Chim.) or with wild-type pHSF1 (HSF1wt). Replicate cultures were then either left at 37°C (Ϫ) or heat-shocked at 43°C for 60 min (HS). Cultures were then rinsed in ice-cold phosphate-buffered saline and lysed in extraction buffer before GST pull-down assay as described under "Materials and Methods." 14-3-3⑀-GST recovered from the lysates by GSH-agarose affinity resin was analyzed by 10% SDS-PAGE. C, I, eluates from the 14-3-3⑀-GST were then analyzed for the presence of HSF1 and the chimera by immunoblot assay using anti-HSF1 or anti-HSF2 antibodies (for HSF1/HSF2). Whole cell extracts (C, II) were also probed for the presence of HSF1 and HSF1/HSF2. Experiments were carried out three times with similar results. D, the 14-3-3⑀-GST fusion protein or GST was immobilized on GSH-linked-Sepharose and incubated with untreated (Ϫ) and heat-shocked (43°C, 1 h) (HS) 35 S-labeled in vitro translated HSF1wt. GST pull-down assay was carried out as described under "Materials and Methods." Samples were analyzed by 10% SDS-PAGE and x-ray film autoradiography. GST was used as a nonspecific binding control. The input lanes represent 10% of the total 35 S-labeled protein used in the assays. E, 35 S-labeled, in vitro-translated HSF1 was either untreated (Ϫ) or exposed to heat shock at 43°C for 1 h (HS) and then incubated either without or with ERK2 and/or GSK3␣ at 30°C for 30 min. GST pull-down assay was performed by incubating in vitro translation mixture with 14-3-3⑀-GST or GST control beads, and complexes were analyzed as described in D. F, 35 S-labeled in vitro translated HSF1 (wt) or HSF1 S303A/307A (mut) was exposed to heat shock, as in E, and then incubated with ERK2 and GSK3␣ at 30°C for 30 min. GST pull-down assay was performed by incubating in vitro translation mixture with 14-3-3⑀-GST or GST control beads, and complexes were analyzed as described in D. Experiments in D-F were each carried out twice with similar results .   FIG. 4. Effects HA-ERK1 overexpression on the association of HSF1 with 14-3-3⑀-GST. Cells were stably transfected with p14-3-3⑀-GST and then transiently transfected with empty vector pcDNA3.1(Vector) or pHA-ERK1. Cultures were then incubated for 24 h, rinsed in ice-cold phosphate-buffered saline, and lysed in extraction buffer before GST pull-down assay as described under "Materials and Methods." 14-3-3⑀-GST recovered from the lysates by GSH-agarose affinity resin was analyzed by 10% SDS-PAGE. I, eluates from the 14-3-3⑀-GST were then analyzed for the presence of HSF1 by immunoblot assay using anti-HSF1 antibodies. Control 14-3-3⑀-GST levels were analyzed using anti-GST antibodies. Total extracts (II) were also probed for the presence of HSF1 and HA-ERK. Experiments were carried out three times with similar results. the HSF1 packed into brightly staining foci of unknown function (Fig. 6A). HA-14-3-3⑀ overexpression at levels that cause more than 90% cytoplasmic sequestration in unstressed cells (24) had little effect on this distribution, and HSF1 in these heat-shocked cells remained more than 80% nuclear and concentrated into the stress foci (Fig. 6B). Our experiments indicate, therefore, that although HSF1 binds tightly to 14-3-3⑀ immediately after heat shock, 14-3-3⑀ association does not lead to cytoplasmic sequestration (Fig. 6) and does not cause transcriptional repression of HSF1 (Fig. 5).
Heat shock is thus able to override the short-term influence of 14-3-3⑀ on cytoplasmic localization, and HSF1 migrates to nuclear sites where it can carry out hsp gene transcription and possible other functions. However, because previous studies suggested a role for ERK phosphorylation in the recovery of cells from heat shock (44) and in the diffusion of HSF1 from nuclear sites, we next examined the effect of 14-3-3⑀ overexpression on HSF1 distribution in the recovery stage after heat shock (Fig. 6C). At 2 and 4 h after heat shock, there was progressive cytoplasmic sequestration of HSF1 in cells overexpressing HA-14-3-3⑀ (Fig. 6C). In controls, HSF1 returned to its cytoplasmic/diffuse nuclear distribution by 4 h after heat, whereas 14-3-3⑀ overexpression led to sequestration of over 80% of the HSF1 in the cytoplasm. DISCUSSION Activation of ERK and induction of HSF1 by heat shock are both important mechanisms in the generation of the stress response, and inhibition of either leads to marked sensitivity to cell stresses (4,10). Their relative interactions seem to be complex. ERK activation and HSF1 hyperphosphorylation (an event associated with transcriptional activation) are closely correlated (Fig. 1). However, ERK1 is a repressor of HSF1 both under growth conditions and in stress (5,6,24). We have characterized the ERK-HSF1 interaction under mitogenic stimulation of G 0 cells, and this interaction involves HSF1 phosphorylation on serine 307 by serum-activated ERK1, secondary HSF1 phosphorylation of primed HSF1 by GSK3 on serine 303, and subsequent association of 14-3-3⑀ with phospho-Ser-303/307-HSF1 (24). This pathway then leads to cytoplasmic sequestration of HSF1, inhibition of HSF1 binding to DNA, and a block to HSP molecular chaperone gene transcription (24). These effects of serum stimulation in G 0 cells are to some extent recapitulated by heat shock (on either G 0 or cycling cells), which leads to prolonged activation of ERK, ERK1 association with HSF1 and ERK-mediated 14-3-3⑀ binding to HSF1 (Figs. 1-4).
Our current studies indicate that heat shock leads to the stable binding of ERK 1 to HSF1 after heat stress and that this interaction may be an initial step in the heat-induced association of HSF1 with 14-3-3 ( Figs. 2 and 3). HSF1 binding to ERK seems to lead to enhanced ERK kinase activity, suggesting a "molecular scaffold" role for HSF1 in ERK activation after stress (Fig. 2B). Indeed, pp-ERK1/2 levels increased significantly in the cytoplasm of heat-shocked cells after HSF1 overexpression. Activation did not involve complexing of the upstream kinases in the ERK cascade with HSF1 and we did not detect either c-raf or MEK2 in HSF1-ERK1 complexes. 3 pp-ERK1/2 levels and elevated ERK kinase activity also increase in untransfected cells after heat shock, suggesting that mechanisms in addition to HSF1 association may also be operative in heat-shocked cells ( Figs. 1 and 2, B and C). Indeed, a number of other mechanisms influence ERK activity in heat-shocked cells, and HSF1-ERK association is likely to be one of a number of complementary mechanisms of ERK activation by heat shock (45). ERK may influence the function of HSF1 as well as the reverse interaction discussed above, and overexpression of ERK led to HSF1-14-3-3⑀ association in accordance with our earlier studies (Fig. 4) (24). We have not determined the exact docking site for ERK1 on HSF1 in this study. However, mutation of serines 303 and 307 inhibited ERK-HSF1 interaction and HSF1 stimulation of ERK kinase activity, suggesting a role for the region encompassing these residues in HSF1-ERK1 binding (Figs. 2, A-C). However, this region in HSF1 is not similar to ERK docking sites characterized in studies of other ERK binding proteins (46,47). The region in ERK that binds to HSF1 is not known, although previous studies indicate that multiple sites in ERK are involved in interaction with other substrates such as MKP3 (48). Another possible mechanism could also involve indirect ERK binding to HSF1 through a bridging interaction with 14-3-3 proteins that bind to the phospho-serine 303/307 region of HSF1 (24). However, previous studies have not identified ERK1 as a 14-3-3 binding protein, and our studies show that 14-3-3 overexpression does not increase HSF1 binding to ERK1-GST, suggesting that this is not a major mechanism for HSF1-ERK interaction. 3 Indeed, when we probed GST-ERK associated proteins, no evidence of 14-3-3⑀ binding was uncovered, indicating that ternary complexes between HSF1, ERK1, and 14-3-3⑀ do not form in any great abundance. Because serines 303 and 307 are needed for binding of both proteins to HSF1, competition between ERK1 and 14-3-3⑀ for HSF1 binding might be predicted (Figs. 2 and 3). In addition, pCMV-lacZ plasmid was co-transfected into each culture as an internal control for transfection efficiency. Cultures were then either left untreated as controls or heat-shocked for 1 h at 43°C before incubation at 37°C to permit luciferase expression. After the experiment, cells were quenched, and proteins were extracted and assayed for expression of transfected HA-14-3-3⑀ and endogenous HSF1. Transfected HA-14-3-3⑀ and endogenous HSF1 were detected by Western blot analysis with anti-HA and specific anti-HSF1 antibodies, respectively, as described under "Materials and Methods." B, relative luciferase activity was next examined in the extracts. Luciferase and ␤-galactosidase were assayed in triplicate samples as described under "Materials and Methods." Luciferase activity in each extract was then normalized to ␤-galactosidase transfection efficiency control activity. Relative luciferase activity was then expressed Ϯ S.D. as a percentage of the activity in cells co-transfected with pGLHSP70B and HA-14-3-3⑀. Luciferase activity was not detected in control cells not transfected with reporter plasmid, and this control was not included in the figure. Experiments were carried out three times with similar results. likely mechanism might involve sequential interactions, with ERK binding and phosphorylation of serine 307 on HSF1 preceding and priming 14-3-3 binding (Fig. 4).
Heat shock leads to HSF1 binding to 14-3-3⑀, and this interaction seems to involve previous ERK association and HSF1 phosphorylation on serine 307, as indicated by our previous studies on serum-stimulated cells (Figs. 2 and 4) (24). However, HSF1-14-3-3⑀ association seemed to be more pronounced after heat shock compared with serum stimulation (Fig. 3) (24). We have previously characterized the region of HSF1 containing phospho-serines 303 and 307 as a major 14-3-3 interaction site, and this is apparently also the case after heat shock (Fig. 3) (24). However, HSF1-14-3-3⑀ association after heat shock was not completely inhibited by serine-alanine mutations of Ser-303 and Ser-307, suggesting the exposure of additional 14-3-3 binding site(s) after heat shock in addition to the serine 303/ 307 region (Fig. 3). This interpretation is also suggested by the in vitro findings that indicate a powerful influence of heat shock on HSF1-14-3-3 interactions (Fig. 3D) but also shows that serines 303 and 307 are required for high affinity binding (Fig. 3F). Heat shock is known to cause the opening up of the HSF1 structure from a highly compacted intramolecular triple stranded coiled-coil into a more extended conformation that forms homotrimers, and this conformational change could potentially expose the serine 303/307 region as well as novel 14-3-3 interaction sites that are cryptic at normal temperatures (49,50). Binding to other molecules, such as NF-IL6 and HSF2A, occurs exclusively after heat shock, indicating the importance of unmasking effects on promoting the association of HSF1 with other proteins (51,52). Although we have not experimentally demonstrated the binding site for HSF1 within 14-3-3⑀, 14-3-3 proteins seem to interact with their substrates through a binding groove conserved in each 14-3-3 subtypes, and it is assumed that HSF1 binds within this structure in 14-3-3⑀ (39). Our previous studies suggest indicate that the phospho-serine 303/307 region of HSF1 binds to a domain in 14-3-3⑀ similar to that bound by the well characterized 14-3-3 substrate CDC25C, good circumstantial evidence indicating that the common ligand binding groove in 14-3-3 is involved in HSF1 binding (24).
Binding to 14-3-3 frequently leads to the cytoplasmic sequestration of client proteins and inhibition of their nuclear functions (37,53,54). This was the finding in the case of HSF1 in serum-stimulated cells (24). However, in heat-shocked cells, 14-3-3⑀ association does not drive HSF1 into a cytoplasmic localization; indeed, HSF1 is almost exclusively nuclear after heat shock (Fig. 6). Because HSF1 binds avidly to 14-3-3⑀ after heat shock, these findings suggest that heat shock overrides the regulatory effects of 14-3-3⑀ operative at normal temperatures (Fig. 3). Our previous studies indicated that 14-3-3⑀ association enhances the interaction of HSF1 with the nuclear export protein CRM1 and leads to increased nuclear export (24). HSF1 has been shown to contain at least two canonical nuclear localization sequences located at the extremes of the FIG. 6. Subcellular localization of HSF1 in cells overexpressing 14-3-3⑀ in response to heat shock. A, HeLa cells were transfected with pHSF1wt-GFP without (1,2) or with pHA-14-3-3⑀ (3). After 48 h of transfection, cells were subjected to heat shock at 43°C for 1 h. Then the cells were fixed and incubated with the primary antibodies anti-14-3-3⑀ or anti-HA. Primary antibody staining was detected with a Texas Redconjugated secondary antibody to localize endogenous 14-3-3⑀ (2) or co-expressed HA-14-3-3⑀ (3) as indicated. HSF1wt-GFP and 14-3-3⑀ were visualized by green autofluorescence or red immunofluorescence as indicated. Cell nuclei were stained by 4,6-diamidino-2-phenylindole autofluorescence. Merge 1 indicates co-localization of green or red fluorescence and the 4,6-diamidino-2-phenylindole (DAPI)-stained nuclei. Merge 2 shows co-localization of HSF1wt-GFP and HA-14-3-3⑀. Whole-cell morphology was visualized in the phase contrast images as indicated. Magnification is 400ϫ. The experiment was carried out three times with reproducible findings. B, quantitative analysis of the subcellular localization of HSF1wt-GFP in HeLa cells co-expressing HA-14-3-3⑀ after heat shock (43°C, 1 h) (HS), as determined by fluorescence microscopy. The subcellular distribution of HSF1wt-GFP was scored according to whether it was predominantly expressed in the nucleus (N), in both the nucleus and the cytoplasm (N & C), or predominantly in the cytoplasm (C). Results are the mean Ϯ S.D. from three separate experiments. C, quantitative analysis of the subcellular localization of HSF1wt-GFP in HeLa cells co-expressing HA-14-3-3⑀ after 1 h at 43°C (HS) followed by 2 or 4 h recovery at 37°C, as determined by fluorescence microscopy. The experiment was performed as described in B. Values shown in the histograms are the mean Ϯ S.D. from three separate experiments. leucine zipper sequence in the N terminus (7) and a number of consensus nuclear export sequences that have so far not been characterized in molecular studies. 4 HSF1 seems to distribute between cytoplasm and nucleus, and its exact destination may depend on the outcome of the competing processes of nuclear import, nuclear export, and cytoplasmic retention (24,55). The function of the bipartite nuclear localization sequence was examined in HSF2 and was shown to be regulated by intramolecular unmasking through an inhibitory leucine zipper domain (LZ4) in the C terminus (56). Masking was relieved by the stress-induced severing of the intramolecular interactions and consequent exposure of the nuclear localization sequence (12,56). One hypothesis to explain the ability of heat shock to override the effects of 14-3-3⑀ on HSF1 localization is that the newly unmasked NLS1 and NLS2 switch the balance toward nuclear localization (Fig. 5). This mechanism does not, however explain the completeness of the switch in properties of the HSF1-14-3-3⑀ complexes (Fig. 6). Another contributing factor could be that HSF1 binds at high affinity to sites in the nucleus that might include gene promoters and stress foci (17). Tethering in the nucleus along with the unmasking of nuclear localization sequences in HSF1 could thus contribute to the completeness of the switch in HSF1-14-3-3⑀ localization from a cytoplasmic localization (24) to a nuclear localization after heat (Fig. 6). Early studies suggested protein tethering to nuclear structures during heat shock (57). Other mechanisms are also suggested in the literature. Heat shock causes the sumoylation of HSF1 on the residue K298 immediately adjacent to the 14-3-3⑀ binding site and this modification may alter the properties of 14-3-3⑀-HSF1 complexes (34,58). In addition, we have found that heat shock affects 14-3-3 proteins, promoting 14-3-3⑀-HSP90 binding and changing the electrophoretic mobility of a minor 14-3-3 species. 2 This suggests that 14-3-3 proteins may not respond in an exclusively passive manner to changes in client phosphoproteins but may also respond directly to stress.
Recovery from stress is associated with the renewed ability of 14-3-3⑀ to influence HSF1 localization to the cytoplasm (Fig.  6C). Thus, one function of HSF1-14-3-3⑀ association may be to aid in the resetting of the stress response during recovery and the return to a non-stress localization. Indeed, earlier studies showed that overexpression of ERK1 or GSK3, enzymes that promote 14-3-3⑀ association with HSF1, accelerate the dissociation of nuclear stress foci containing HSF1 and promote the attenuation of the stress response (32). Thus, through stable interaction with 14-3-3⑀, HSF1 is poised to recover its cytoplasmic localization once the acute phase of the response to stress is over. Because HSF1 activity and HSP expression are antagonistic to cell growth, efficient termination of the heat shock response may be crucial for the resumption of growth and the survival of cell populations (59,60). It is not clear whether all the effects of ERK-HSF1 association are exerted through hsp gene transcription. Indeed, because ERK inhibits HSP synthesis yet is essential for thermotolerance, this seems unlikely (4, 5). 14-3-3 proteins play a key role in antagonizing apoptosis (61)(62)(63). Recruitment of 14-3-3 by HSF1 may participate in this effect, and our preliminary studies indicate that 14-3-3 dominant-negative constructs inhibit HSF1-mediated thermotolerance. 5