Regulation of Molecular Chaperone Gene Transcription Involves the Serine Phosphorylation, 14-3-3ɛ Binding, and Cytoplasmic Sequestration of Heat Shock Factor 1

Abstract

Heat shock factor 1 (HSF1) regulates the transcription of molecular chaperone hsp genes. However, the cellular control mechanisms that regulate HSF1 activity are not well understood. In this study, we have demonstrated for the first time that human HSF1 binds to the essential cell signaling protein 14-3-3ɛ. Binding of HSF1 to 14-3-3ɛ occurs in cells in which extracellular signal regulated kinase (ERK) is activated and blockade of the ERK pathway by treatment with the specific ERK pathway inhibitor PD98059 in vivo strongly suppresses the binding. We previously showed that ERK1 phosphorylates HSF1 on serine 307 and leads to secondary phosphorylation by glycogen synthase kinase 3 (GSK3) on serine 303 within the regulatory domain and that these phosphorylation events repress HSF1. We show here that HSF1 binding to 14-3-3ɛ requires HSF1 phosphorylation on serines 303 and 307. Furthermore, the serine phosphorylation-dependent binding of HSF1 to 14-3-3ɛ results in the transcriptional repression of HSF1 and its sequestration in the cytoplasm. Leptomycin B, a specific inhibitor of nuclear export receptor CRM1, was found to reverse the cytoplasmic sequestration of HSF1 mediated by 14-3-3ɛ, suggesting that CRM1/14-3-3ɛ directed nuclear export plays a major role in repression of HSF1 by the ERK/GSK3/14-3-3ɛ pathway. Our experiments indicate a novel pathway for HSF1 regulation and suggest a mechanism for suppression of its activity during cellular proliferation.

Heat shock transcription factor 1 (HSF1) is the mammalian regulator of the heat shock response and activates the transcription of heat shock protein (Hsp) molecular chaperone genes (43, 47, 49). Inactivation of the murine hsf1gene has been shown to confer a complex phenotype, indicating an essential function for hsf1 in growth, in development, and in acute response to stress (35). Disruption of hsf1 (i.e., hsf1^−/−) in mouse embryonic fibroblasts leads to a profound loss of thermotolerance and markedly increased susceptibility to heat-induced apoptosis (11, 35). hsf1 is required as a maternal factor during the early cleavage stage of development in the ^−/− mouse embryo (11). hsf1-deficient mice can survive to adulthood but display severe defects in the chorioallantoic placenta that result in increased prenatal lethality (56). In addition, the aging process is associated with degeneration of the heat shock response, and transcriptional activity of HSF1 protein was significantly reduced with age in a cell-free system, as well as in isolated hepatocytes (26). Understanding the processes involved in HSF1 regulation may therefore aid in delineating its role in resistance to stress, development, and aging.

Under normal conditions, cellular HSF1 exists in a predominantly transcriptionally repressed state (44, 59). Such HSF1 is monomeric, is constitutively phosphorylated, and lacks the ability to bind the cis-acting heat shock elements (HSEs) located in the promoters of Hsp genes (50, 55). Induction of transcriptional activity by heat shock then results in the conversion of HSF1 from inactive monomer to a DNA-binding trimer (44, 46, 53). Activation of HSF1 is a multistep process, involving trimerization, acquisition of HSE-binding activity, novel phosphorylation, and transactivation of Hsp genes (38, 50, 55). 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 (44). The molecular chaperone Hsp90 functions as the principal cellular 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 (59). Hsp70, is also important, but not sufficient, to negatively regulate HSF1 activity in the absence of stress (1, 5). HSF1 also appears to have additional layers of regulation in the cell (25, 42). Our studies, as well as the work of others, have demonstrated the hierarchical phosphorylation of human HSF1 within its transcriptional regulatory domain by extracellular signal-regulated kinase 1 (ERK1; on serine 307) and by glycogen synthase kinase 3 (GSK3; on serine 303) (12, 13, 23, 29, 30). The regulatory domain of HSF1 functions as a molecular switch coupling hsp gene transcription to cellular conditions, repressing C-terminal transactivation domains under growth conditions, and under stress conditions causing powerful stimulation of the same activation domains (22, 30, 40). In the present study, we have examined how the regulatory domain responds to extracellular conditions. Since phosphorylation can directly regulate distinct aspects of transcription factor function, including subcellular localization, protein stability, protein-protein interactions, DNA binding, and transactivation, we have examined the role of phosphorylation in the activity of the regulatory domain (14, 54).

Our previous studies indicate that the protein kinases ERK1 and GSK3 phosphorylate HSF1 on serine residues within a proline-rich region (RVKEEPPS³⁰³PPQS³⁰⁷PRV) of the regulatory domain (12, 13). Many recent studies show that phosphorylation on serine and threonine can be converted into an intracellular signal by association of the phosphorylated domain with regulatory proteins that recognize serine/threonine phosphorylated domains (14). The first such proteins to be identified were the highly conserved 14-3-3 family, which bind to a wide array of cellular proteins largely through recognition of phosphoserine-containing domains (2, 57). At least seven 14-3-3 genes exist in vertebrates, and these give rise to nine protein isoforms (α, β, δ, ɛ, γ, η, σ, τ, and ζ) (3). 14-3-3 proteins are found largely as dimers within the cell and are able to bind either to multiple sites within proteins such as c-Raf1 or to act as a bridge, with one 14-3-3 dimer binding to two different proteins (20). 14-3-3 dimers can thus act as molecular scaffold proteins, bringing together proteins that interact functionally and effecting phosphorylation-dependent cell regulation (20). These highly abundant proteins are thus involved in key cellular processes, such as signal transduction, cell cycle control, and apoptosis (20, 34, 57, 58).

In the present study, we investigated the potential role of 14-3-3 in the function of the transcriptional regulatory domain of HSF1. We show for the first time that HSF1 can bind to 14-3-3. In addition, we have demonstrated that extracellular ERK activation by mitogenic stimulation, a process that controls phosphorylation of serines 303 and 307 of HSF1, leads to HSF1 association with 14-4-3 species, including 14-3-3ɛ and 14-3-3ζ. In addition, inhibiting the ERK cascade with the specific MEK1 inhibitor PD98059 in vivo strongly suppresses the binding of HSF1 to 14-3-3ɛ. We next demonstrated the direct in vitro binding of 14-3-3ɛ to synthetic HSF1-derived peptides phosphorylated at serines 303 and 307. Our studies further showed that 14-3-3ɛ binding inhibits both the transcriptional activity and the nuclear accumulation of HSF1 in HeLa cells at 37°C. These effects of 14-3-3ɛ on nuclear exclusion and transactivation of HSF1 required serines 307 and 303 within the regulatory domain. Leptomycin B (LMB), specific inhibitor of nuclear export receptor CRM1, blocked the cytoplasmic localization of HSF1 in 14-3-3ɛ-overexpressing cells. These experiments suggest that 14-3-3ɛ mediates interaction between phospho-S307-S303-HSF1 and CRM1 in either a direct or an indirect manner and thus contributes to cytoplasmic sequestration and transcriptional repression in response to ERK stimulation. Our experiments therefore indicate a pathway whereby HSF1 activity and molecular chaperone expression are regulated at the intracellular level by protein kinases and within larger biological systems through the mediation of extracellular factors.

MATERIALS AND METHODS

Plasmid constructions and mutagenesis.

The human HSF1 full-length cDNA was amplified by using specific PCR primers and subcloned into pHM6 (Roche) to generate pHM6HSF1wt. This construct was used for site-directed mutagenesis by using the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. Two pairs of primers (5′-GAGGAGCCCCCCGGCCCGCCTCAGA-3′ and 3′-CTCCTCGGGGGGCCGGGCGGAGTCT-5′; 5′-CGCCTCAGGGCCCCCGGGTAGAGGA-3′ and 3′-GCGGAGTCCCGGGGGCCCATCTCCT-5′) were used in mutating serine residues at 303 (pHM6S303G) and 307 (pHM6S307G) to glycine. To construct HSF1-GFP fusion proteins, cDNA of human wild-type and mutants pHM6HSF1 were fused to the green fluorescent protein (GFP) sequence in pEGFP-N₃ (Clontech) at HindIII-EcoRI sites and are named HSF1wt-GFP, S303-GFP, and S307-GFP. S303G-GFP or pcDNA3.1 S303G was used to construct the double mutants S303G-S307G-GFP or pcDNA3.1 S303G-S307G by using the same pair of primers as that used for pHM6S307G. All constructs and mutations were confirmed by sequencing. HSP70B promoter luciferase reporter construct (pGLHSP70B) and other expression plasmids of HSF1 wild type and mutants (pcDNA3.1 HSF1wt, pcDNA3.1 S303G, pcDNA3.1 S307G, and pcDNA3.1 S363A) have been described previously (10, 12).

Cell culture and transient transfection.

HeLa cells were cultured in Ham F-12 (Mediatech, Inc.) supplemented with 10% heat-inactivated fetal bovine serum (FBS). HeLa cells (2.5 × 10⁵ cells/well) in six-well plates were transfected with the plasmids indicated in the figure legends in triplicate by using FUGENE6 (Roche) according to the manufacturer's protocol as described previously (52). In transient transfection, wild-type or mutant pcDNA3.1 HSF1 (100 ng/well) was cotransfected with pGLHSP70B promoter luciferase reporter plasmid at 400 ng/well or with hemagglutinin (HA)-tagged 14-3-3ɛ expression vector, pcDNA3.1 HA-14-3-3ɛ at 500 ng/well (HA-14-3-3ɛ was a kind gift from Sorab N. Dalal, Dana-Farber Cancer Institute, Boston, Mass.) (Fig. 1B). Other amounts of HA-14-3-3ɛ are as indicated in the figures. pCMV-lacZ plasmid was cotransfected 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 β-galactosidase activity assays were then performed according to the Promega protocol. Luciferase activity was normalized to β-galactosidase activity. The results were expressed as the relative luciferase activity of the appropriate control.

FIG. 1.

Overexpression of 14-3-3ɛ represses transcriptional activation of the HSP70B promoter by endogenous HSF1. HeLa cells were transfected with pGLHSP70B (HSP70B) alone or with increasing amounts of HA-14-3-3ɛ expression plasmid as indicated. In addition, pCMV-lacZ plasmid was cotransfected into each culture as an internal control for transfection efficiency. (A) After each experiment, cells were quenched and proteins were extracted and assayed for expression of transfected HA-14-3-3ɛ and endogenous HSF1. Transfected 14-3-3ɛ was detected by Western analysis with anti-HA antibodies and HSF1 with specific anti-HSF1 antibodies as described in Materials and Methods. (B) The relative luciferase activity was next examined in the extracts. Luciferase and β-galactosidase were assayed in triplicate samples as described in Materials and Methods. The luciferase activity in each extract was then normalized to the β-galactosidase transfection efficiency control activity. The relative luciferase activity was then expressed ± the standard deviation (SD) of the mean as a percentage of the activity in cells cotransfected with pGLHSP70B and HSF1wt (second column). 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. (C) Plot of the total activity of the HSP70B-luciferase promoter and the pCMV-lacZ control promoter (determined as the β-galactosidase activity) calculated in the experiments above, illustrating the lack of effect of 14-3-3ɛ overexpression on the activity of the control promoter.

Protein interaction in vivo.

To facilitate detection of 14-3-3-HSF1 complexes formed in vivo, 14-3-3ɛ was expressed in cells as a glutathione S-transferase (GST) fusion protein through a modified pEF1-BOS vector, and bound proteins were analyzed by using glutathione (GSH)-linked Sepharose pull-down assays as previously described (21). Briefly, cells were transfected with a total of 0.5 μg of DNA by using Lipofectamine (Invitrogen). After 24 h, cells were incubated in fresh complete medium or serum starved, depending on the experiment. Cells were then lysed in extraction buffer consisting of 50 mM HEPES (pH 7.5), 0.5% Triton X-100, 2 mM EGTA, 5% glycerol, 50 mM glycerophosphate, and 1 mM dithiothreitol plus protease and phosphatase inhibitors (1 mM sodium vanadate, 1 mM NaF, 0.4 mM phenylmethylsulfonyl fluoride, and 10 μg each of leupeptin and aprotinin/ml). For immunoprecipitations under stringent conditions, the extraction buffer was supplemented with 0.1% sodium dodecyl sulfate (SDS) and 0.5 M NaCl. 14-3-3ɛ-GST fusion proteins were purified by using preequilibrated GSH-Sepharose, and incubations were performed for 16 h at 4°C (21). After three washes with extraction buffer, GSH-Sepharose-bound proteins were eluted with Laemmli SDS-loading buffer (25 mM Tris [pH 7.0], 1% SDS, 2.5% mercaptoethanol, 0.5 mg of bromophenol blue/ml, 5% glycerol), separated by SDS-12.5% polyacrylamide gel electrophoresis (PAGE), and transferred to a Hybond-ECL membrane (Amersham Biosciences, Inc.). Immunoblot detection was performed with the antibodies dissolved in 5% dried skim milk in phosphate-buffered saline and developed by using alkaline phosphatase-coupled secondary antibodies and an enhanced chemiluminescence system according to the manufacturer's instructions (Amersham Biosciences, Inc.). Antibodies for GST (Z-5), 14-3-3ɛ (SC-1020), 14-3-3 (pan-14-3-3 and K-29), and ERK2 (SC-1647) were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, Calif.) For detection of HSF1 a mixture of anti-HSF1 specific antibodies, consisting of A68-3 (12), SPA-950 (Stressgen), and SC-13516/8061 (from Santa Cruz) was used. Anti-active ERK1/2 and MEK polyclonal antibodies were from Biosource International. For reblotting, membranes were incubated in 25 mM dithiothreitol-1.5% SDS in phosphate-buffered saline for 20 min at an ambient temperature.

For the experiment described in Fig. 2C, 14-3-3ɛ-GST was expressed in COS-1 cells and purified under stringent conditions as described above. The lack of associated polypeptides, including endogenous HSF1, was confirmed by Coomassie blue staining and by Western blotting. 14-3-3ɛ-GST, still bound to the GSH-Sepharose beads, was mixed with 200 μl of extract from cells stimulated with serum as described above. These extracts involved cultures that had been prepared by transfection with FLAG-HSF1 or vector alone, followed by stimulation with 20% fetal calf serum. The mixtures described above were incubated at 4°C for 2 h with or without HSF1 S303 phosphorylated peptide (pS303) or HSF1 S307 phosphorylated peptide (pS307), the latter added at 200 μg/ml each. The 14-3-3ɛ-GST Sepharose beads were washed three times in the extraction buffer described above and then eluted in Laemmli SDS-loading buffer. Associated FLAG-HSF1 was detected by Western blotting with anti-FLAG (M5) antibodies.

FIG. 2.

HSF1 binds to 14-3-3 after serum stimulation and requirement for HSF1 phosphorylation on serines 303 and 307. (A) Association of 14-3-3 with HSF1 is enhanced by serum. GST-tagged 14-3-3ɛ expression vector was transfected in equivalent amounts into four subconfluent cell cultures, and plates were serum starved as described in Materials and Methods. Subsequently, the cultures were harvested without stimulation (lanes 1 and 2) or after stimulation with 20% fetal calf serum. 14-3-3ɛ-GST fusion proteins were then purified by GSH affinity chromatography, fractionated by SDS-PAGE, and analyzed for their levels and associated proteins with anti-GST and anti-HSF1 antibodies as indicated. As a control, extracts from the two serum-stimulated duplicate cultures of lanes 3 and 4 were subjected to GSH affinity chromatography in the presence of 0.1% SDS-0.5 M NaCl (lanes 5 and 6). As additional control pull-downs, protein A-Sepharose was used in place of GSH-Sepharose for these two extracts. Finally, total extracts from each of the four transfections were analyzed by Western and enhanced chemiluminescence with antibodies against activated MEK or ERK or with control antibodies for ERK2 or HSF1 (bottom). For direct assessment of the relative levels of 14-3-3ɛ-GST expression, the same extracts were also blotted with anti-pan 14-3-3 antisera, which recognizes both the transfected 14-3-3ɛ-GST chimera and the endogenous 14-3-3 isoforms. To this end, please note the low levels of exogenous 14-3-3 expression in comparison with its endogenous counterpart. Experiments were carried out twice with reproducible results. (B) Association of 14-3-3 with HSF1 in vivo. 14-3-3ɛ was transiently expressed in cells as a GST fusion protein as described in Materials and Methods. As a control, in a parallel transfection, GST propeptide was expressed alone as indicated. Cycling cultures of each of the three transfections were harvested, solubilized in extraction buffer, and subjected to GSH affinity chromatography (lanes 1, 3, and 5). As controls, parallel pull-downs containing 0.1% SDS-0.5 M NaCl in the extraction and wash buffer were included (lanes 2, 4, and 6). After SDS-PAGE, anti-GST and anti-HSF1 blotting was performed to verify the levels of 14-3-3ɛ-GST expression and recovery and associated HSF1. For assessment of protein expression, control Western blots of total cellular extracts with anti-pan 14-3-3 antisera or antibodies directed against HSF1 are also shown (bottom panels). (C) HeLa cells were serum starved and pretreated with 50 μM PD98059 before stimulation with 10% FBS-Ham F-12 medium. Cells were then lysed in immunoprecipitation buffer and probed by immunoprecipitation with anti-HSF1 polyclonal antibody (A68-3), followed by immunoblotting with anti-14-3-3ɛ monoclonal antibody as indicated (upper panel). The blot was then reprobed to confirm efficient immunoprecipitation of HSF1 with anti-HSF1 polyclonal antibody (A68-3) (bottom panel). Experiments were carried out twice with reproducible findings. (D) Binding of ³⁵S-labeled 14-3-3ɛ to synthetic peptides. Microtiter wells were coated with synthetic phosphorylated and unphosphorylated peptides derived from the proline-rich domain of HSF1 (phospho-S303-S307, phospho-S307, phospho-S303, and unphospho-HSF1) as described in Materials and Methods. Different concentrations of ³⁵S-labeled 14-3-3ɛ protein (160 and 400 ng/ml) were then added to the coated microtiter wells and incubated at 22°C for 2 h. After an extensive washing, bound proteins were extracted and measured by liquid scintillation counting. The results are expressed as the mean of the ³⁵S-labeled 14-3-3ɛ activity extracted from triplicate wells ± the SD. Experiments were carried out three times with close agreement between experiments. (E) Competitive binding of phosphorylated HSF1 peptides to ³⁵S-labeled 14-3-3ɛ protein. ³⁵S-labeled 14-3-3ɛ (160 ng/ml) was incubated first with a 100 μM concentration of each of the phosphorylated or unphosphorylated peptides for 1 h at 22°C and then added to microtiter wells coated with each of the phosphorylated peptides (phospho-S303-S307, phospho-S307, and phospho-S303) (as described above), followed by incubation at 22°C for 2 h. After extensive washing of the plates, bound proteins were extracted and assayed by liquid scintillation counting. Competition experiments were also carried out with phospho-Cdc25c-S216, a well-characterized 14-3-3-binding peptide, as a positive control as indicated. Each competition assay was carried out in duplicate. The entire experiment was carried out three times with reproducible findings each time. (F) Effect of phospho-S303 and phospho-S307 peptides on the association of 14-3-3 with FLAG-HSF1. Purified 14-3-3-GST was mixed with serum-stimulated cell extracts prepared from cells expressing either FLAG-HSF1 (lane 1) or FLAG propeptide alone (lane 4). To test the potential effect of phosphorylation on the association of 14-3-3ɛ with FLAG-HSF1, parallel reactions included synthetic HSF1-based peptides phosphorylated either on Ser-303 (lanes 2 and 5) or Ser-307 (lanes 3 and 6). After washes, the 14-3-3ɛ-GST and associated FLAG-HSF1 were fractionated by SDS-PAGE and blotted with anti-FLAG and anti-14-3-3ɛ antibodies as indicated. Below each lane the densitometry value of bound FLAG-HSF1 is indicated as a percentage value, with a control (100%) in lane 1. As a control, 10-μl aliquots of the reactions taken before washes were also assessed for the levels of FLAG-HSF1 (bottom). Experiments were repeated reproducibly three times.

Immunoprecipitation.

To detect the interaction between HSF1 and endogenous 14-3-3ɛ, HeLa cells were washed once in 0.1% FBS in Ham F-12 medium and then maintained in this medium for 72 h to arrest cells in G₀ (6). Serum-starved cells were pretreated with 50 μM PD98059 (Cell Signaling Technology, Inc.) for 1 h. Control incubations contained vehicle (dimethyl sulfoxide) alone. Cells (at the 72 h-time period) were then induced to enter G₁ by serum refeeding as described previously (6). Cell lysates were prepared from PD98059-treated and untreated HeLa cells by using immunoprecipitation buffer (20 mM HEPES [pH 7.4], 0.5% NP-40, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsufonyl fluoride, 10 μg of aprotinin/ml, 10 μg of leupeptin/ml, 50 mM NaF, 1 mM Na₃VO₄). The lysates (8 mg [total amount of protein]) were precleared with protein A-Sepharose for 30 min at 4°C. Cleared extracts were then incubated with anti-HSF1 polyclonal antibody (A68-3) overnight at 4°C and then with protein A-Sepharose for 2 to 4 h under gentle shaking. The immunoprecipitated materials were washed extensively with immunoprecipitation buffer and analyzed by SDS-10% PAGE, followed by immunoblotting with anti-14-3-3ɛ monoclonal antibody (Transduction Laboratories).

Western analysis.

Cells were lysed and protein concentration was quantified in the cell lysates by using the DC protein assay (Bio-Rad Laboratories, Richmond, Calif.). Cytoplasmic and nuclear extracts were prepared as described previously (48); whole-cell lysates were also prepared as described previously (6). Samples were then subjected to SDS-10% PAGE transferred to polyvinylidene difluoride membranes and subjected to Western analysis as described previously (52). The following antibodies were used in the studies: anti-HSF1 (A68-3) (6), anti-HA (Roche Products), anti-14-3-3ɛ (Transduction Laboratories), and anti-α-actin (Santa Cruz Biotechnology).

14-3-3ɛ-HSF1 phosphopeptide-binding studies.

³⁵S-labeled 14-3-3ɛ protein was produced by in vitro translation in the presence of [³⁵S]methionine (>1,000 μCi/mmol; NEN) according to the manufacturer's protocol by using the T7-TNT Quick-Coupled transcription-translation system (Promega) with pcDNA3.1HA-14-3-3ɛ expression plasmids. Peptides (nonphosphopeptides and phosphopeptides) were synthesized and purified to 99% purity by reversed-phase high-performance liquid chromatography by a commercial vendor (Biosource, Inc.). The peptide sequences were as follows: HSF1303-307 phosphorylated peptide (phospho-S303-S307), acetyl-P(pS)PPQ(pS)PRVC-amide; HSF1303-307 nonphosphorylated peptide (unphospho-HSF1), acetyl-PSPPQSPRVC-amide; HSF1 S303 phosphorylated peptide (phospho-S303), acetyl-P(pS)PPQSPRVC-amide; HSF1-307 phosphorylated peptide (phospho-S307), acetyl-PSPPQ(pS)PRVC-amide; and Cdc25c S216 phosphorylated peptide(phospho-Cdc25c-S216), acetyl-RSP(pS)MPC-amide. 14-3-3 peptide binding was studied using techniques similar to those described previously (17, 39). Briefly, the synthetic peptides were solubilized in distilled water and then diluted to a final concentration in 0.1 M NaHCO₃ (pH 9.2). The peptides were next coated onto microtiter wells (Immunolon 4HBX; Dynatech Laboratories, Chantilly, Va.), followed by incubation at 22°C for 6 h and then at 4°C overnight. The peptide-coated microtiter wells were incubated with 5% bovine serum albumin for 2 h at 22°C and then with ³⁵S-labeled 14-3-3ɛ protein for 2 h. After an extensive wash with ice-cold phosphate-buffered saline, 5 μl of SDS-PAGE sample buffer was added to each well, and then eluted ³⁵S-labeled 14-3-3ɛ was removed and assayed by liquid scintillation counting. Unphosphorylated HSF1 and phospho-Cdc25c-S216 peptides used in control incubations were prepared and purified in a manner similar to that used for the phosphorylated HSF1 peptides.

EMSA for HSF1-HSE binding.

Nuclear extracts were prepared from cells as described in the text, incubated with a double-stranded, ³²P-labeled consensus HSE from human HSP70B promoter probe, and analyzed by electrophoretic mobility shift assay (EMSA) as described previously (52). To prepare control nuclear extracts from heat-shocked cells, HeLa cultures were exposed to 43°C for 1 h, and nuclear extracts were then made as described above.

Intracellular localization of HSF1 and 14-3-3ɛ by fluorescence microscopy.

After transfection of HeLa cells growing on four-chamber tissue culture slides with GFP-HSF1 constructs or HA-14-3-3ɛ or HA-P16 (3 × 10⁴ cells/chamber, 1 μg of DNA of GFP expression plasmid/chamber) cells were fixed with 4% paraformaldehyde in phosphate-buffered saline. For immunofluorescence, fixed cells were incubated with primary antibodies, anti-HA, anti-14-3-3ɛ, anti-HSF1, or anti-HSF2 (Stressgen) antibody, and antibody binding was detected by using Texas red-conjugated secondary antibody (Amersham Pharmacia Biotech) or fluorescein isothiocyanate-conjugated secondary antibody (Jackson Laboratories). In all experiments, cells were then counterstained with DAPI (4′,6′-diamidino-2-phenylindole; Roche) to visualize the nucleus. Images were acquired by using a Nikon Eclipse E600 microscope equipped with an RT color SPOT digital camera and processed by using SPOT software (Diagnostic Instruments, Inc.). To specifically block CRM1-mediated nuclear export, cells were exposed to the inhibitor LMB (Sigma). HeLa cells transfected with the HSF1-GFP constructs were treated without or with LMB (10 ng/ml) for 4, 6, and 16 h prior to fixation.

RESULTS

14-3-3ɛ expression represses the transcriptional activity of HSF1.

We first tested the hypothesis that 14-3-3 proteins could act as intracellular regulators of HSF1-mediated transcription. We overexpressed 14-3-3 in HeLa cells by using a cytomegalovirus (CMV) promoter-based, HA-tagged 14-3-3ɛ expression vector. (14-3-3ɛ was chosen for the availability of reagents; our studies indicate that HSF1 binds at least two 14-3-3 species and perhaps others.) Although 14-3-3 proteins are highly abundant, they appear to be present at critical levels in cells due to the high concentration of binding ligands, and even small increases in 14-3-3 expression, as accomplished here, may mimic physiological regulatory changes mediated by 14-3-3 (20). To examine HSF1 activity, we used an HSP promoter-reporter luciferase construct pGLHSP70B from the highly HSF1 selective HSP70B promoter (9, 12). HeLa cells were transfected with pGLHSP70B alone or together with increasing amounts of HA-14-3-3ɛ expression vector. As can be seen, this led to the expression of increasing concentrations of HA-14-3-3ɛ in a plasmid dose-dependent manner and to HSP70B repression with a similarly HA-14-3-3ɛ-dependent relationship (Fig.1). Expression of HA-14-3-3ɛ did not affect the intracellular levels of HSF1, which remained constant despite the increasing concentrations of HA-14-3-3ɛ (Fig. 1A). These experiments show that overexpression of HA-14-3-3ɛ leads to the repression of endogenous HSF1 activity and implicate HA-14-3-3ɛ in HSF1 regulation (Fig. 1). However, they do not indicate whether these effects are due to direct interaction between HSF1 and HA-14-3-3ɛ as we hypothesize or to indirect or pleiotropic effects of HA-14-3-3ɛ overexpression. Evidence for promoter selectivity in these effects of HA-14-3-3ɛ is provided by examining the activity of the transfection efficiency control pCMV-lacZ promoter-reporter (Fig. 1C). Despite the inhibition of the HSP70B promoter activity, HA-14-3-3ɛ overexpression failed to repress the CMV immediate-early promoter at the concentrations used in these experiments, indicating some selectivity for HSP70B (Fig. 1C).

14-3-3ɛ binds to HSF1 in cells with activated ERK kinases; binding requires HSF1 phosphorylation on serines 303 and 307.

We next investigated whether HA-14-3-3 repression of the HSP70B promoter involves HSF1 binding to HA-14-3-3. Our preliminary studies indicate that HSF1 can interact with a number of 14-3-3 isoforms, as indicated by probing HSF1 immunoprecipitates with pan-13-3-3 antibodies and antibodies specific for 14-3-3ɛ and 14-3-3ζ (N. Grammatikakis and S. K. Calderwood, unpublished results). We have concentrated on the ɛ isoform due to the availability of a range of reagents for study. Cells were transfected with an expression vector that leads to the intracellular expression of a GST-14-3-3ɛ fusion protein. The association of HSF1 with GST-14-3-3ɛ was then determined after cell lysis, GST pull-down, and Western analysis. We examined the potential role of ERK phosphorylation of HSF1 in this process by comparing HSF1-GST-14-3-3ɛ association in quiescent cells in which ERK activity is minimal to mitogenically stimulated cells with strong ERK activation (Fig. 2A). The experiments indicate significant HSF1 association with 14-3-3ɛ in the mitogenically activated cells (Fig. 2AI, lanes 3 and 4) but not in the quiescent cells (Fig. 2AI, lanes 1 and 2). These experiments did not involve the use of overwhelming concentrations of GST-14-3-3ɛ, as can be seen in the control experiment (Fig. 2AII), and the levels of GST-14-3-3ɛ in the transfectants are considerably lower than the levels of endogenous 14-3-3ɛ. The relative levels of activation of the ERK pathway in these samples are indicated in the control experiments (Fig. 2AII). Activation of the MEK1 and ERK1/2 intermediates by phosphorylation, detected by using phosphospecific antibodies, was observed only in the mitogenically stimulated cells in which HSF1-14-3-3ɛ association was observed (Fig. 2AII). Total expression levels of HSF1 and ERK2 are not affected by these experimental conditions, as shown by the control experiments in Fig. 2A. Further control experiments carried out in Fig. 2B show that the HSF1-14-3-3ɛ association is specific, not observed with wild-type GST, and not observed when the pull-down experiments are carried out under denaturing conditions (SDS and high salt). We next used PD98059, a specific inhibitor of the dedicated ERK pathway intermediate MEK-1, to examine the requirement for the ERK pathway in HSF1-14-3-3ɛ association. PD98059 has been shown to act in vivo as a highly selective inhibitor of MEK1 activation and the mitogen-activated protein kinase cascade (18). HeLa cells were serum starved, preincubated with or without 50 μM PD98059, and then induced to enter G₁ by serum refeeding as described previously (6). In these experiments we used an alternative approach to studying HSF1-14-3-3ɛ interaction, in this case involving the immunoprecipitation of native HSF1, followed by immunoblot analysis of the immunoprecipitates with anti-HSF1 and anti-14-3-3ɛ antibodies. We observed that native HSF1 and 14-3-3ɛ are coimmunoprecipitated from cells after serum stimulation (Fig. 2C, lane 2) and that pretreatment of cells with 50 μM PD98059 strongly decreases the association of 14-3-3ɛ with immunoprecipitated HSF1 (Fig. 2C, lane 1). These experiments suggest that activation of ERK is an important factor in HSF1-14-3-3ɛ association after extracellular stimulus (Fig. 2A to C).

We next investigated the hypothesis that the association of HSF1 with 14-3-3 involves direct binding of 14-3-3ɛ to HSF1 phosphorylated at serines 307 and 303 after ERK activation in vivo (12, 13). We examined 14-3-3ɛ binding to synthetic peptides derived from HSF1, which were phosphorylated either at serine 303 or serine 307 during synthesis. Peptides containing HSF1 sequences surrounding serines 303 and 307 of HSF1 (residues 302 to 311) were synthesized either without phosphorylation, singly phosphorylated on Ser-303 or Ser-307, or doubly phosphorylated on Ser-303 and Ser-307 and then purified to >99% purity by reversed-phase high-performance liquid chromatography. These purified phosphopeptides were then used to coat the wells of microtiter plates, as described in previous studies aimed at determining 14-3-3 phosphopeptide binding (17, 39). Plates were then incubated with ³⁵S-labeled 14-3-3ɛ, and relative levels of peptide-14-3-3ɛ binding were examined. ³⁵S-labeled 14-3-3ɛ was bound to the phosphorylated peptides (phospho-S303-S307, phospho-S307, and phospho-S303) in a dose-dependent manner (Fig. 2D). The relative degree of HSF1-14-3-3ɛ-peptide association increased in the following order: phospho-S303 < phospho-S307 < phospho-S303-S307 (Fig. 2 D). In contrast, there was no specific binding of 14-3-3ɛ to the nonphosphorylated HSF1 peptide at a 14-3-3ɛ concentration of 160 ng/ml, although slight binding was observed at 400 ng/ml (Fig. 2D). HSF1-14-3-3ɛ association with the nonphosphorylated peptide is likely nonspecific, since it was not competed for by the addition of any of the phosphopeptides or with nonphosphorylated peptide itself. 14-3-3ɛ binding to the peptides was enhanced by the presence of phosphoserine at either position, and 14-3-3ɛ binding to could be cross-competed with each of the phosphorylated peptides (phospho-S303-S307, phospho-S307, and phospho-S303) but not with unphosphorylated HSF1 peptide (Fig. 2E). We also carried out competition experiments with a known 14-3-3ɛ-binding phosphopeptide from the mitotic phosphatase cdc25c (16), phospho-cdc25c-S216 (Fig. 2E). Phospho-cdc25c-S216 competed effectively for 14-3-3ɛ binding to each of the HSF1 phosphopeptides, suggesting that the cdc25c and HSF1 phosphopeptides may contact a similar site in 14-3-3ɛ (Fig. 2E). Our experiments therefore suggest that 14-3-3ɛ binds directly to HSF1 and that binding requires serine phosphorylation at serines 303 and 307 and, moreover, show that the binding is strongest when both residues are phosphorylated. Since these experiments were carried out with the HSF1 phosphopeptides adsorbed to a solid phase, we next examined the ability of the peptides to interact with 14-3-3ɛ in free solution (Fig. 2F). Cells expressing HSF1-FLAG were serum starved and refed and the phospho-FLAG-HSF1 containing extracts prepared. We then examined the ability of purified GST-14-3-3ɛ to interact with phospho-HSF1-FLAG in solution (Fig. 2F). We observed that GST-14-3-3ɛ became associated with HSF1-FLAG in the extracts and that this interaction was perturbed by incubation with the HSF1 phosphopeptides (Fig. 2F). Phospho-S307 efficiently inhibited this interaction by 50 to 60% in repeated experiments (lane 3), whereas phospho-S303 showed less-effective inhibition with 20% antagonism of HSF1-14-3-3 binding in the three repeated experiments (lane 2), a finding in line with the results of the experiments described above (Fig. 2D to F). As can be seen in the control experiment, similar amounts of HSF1-FLAG were input in each of the incubations, and HSF1-FLAG was absent from incubations with materials from vector control cells, which were not transfected with HSF1-FLAG expression vector (Fig. 2FII).

Repression of HSF1 transcriptional activity by 14-3-3ɛ requires serines 303 and 307.

We next examined the potential role of serines 303 and 307 in the transcriptional repression of HSF1 by 14-3-3ɛ. We coexpressed either wild-type HSF1 or HSF1 in which Ser-303 or Ser-307 were mutated to glycine or alanine. The glycine and alanine mutants had essentially identical phenotypes, and we show here the results of experiments with the glycine mutants for continuity with previous publications (12, 13). Experiments were carried out by overexpression of wild-type HSF1 and HSF1 point mutants S303G and S307G and the double mutant S303G-S307G, in which serine residues 307 and/or 303 were mutated to glycine, resulting in inactivation of phosphorylation by, respectively, ERK1 and/or GSK3 (12, 13). We found that repression of HSP70B promoter activity by 14-3-3ɛ was effectively blocked by mutations at either phosphorylation site (Fig. 3B). Similar results were observed with the corresponding Ser/Ala mutants, suggesting that these effects were due to the loss of phosphorylation of serines 303 and 307. The experiments therefore suggest that the role of 14-3-3ɛ in the repression of HSF1 is dependent on phosphorylation at both serines 303 and 307 (Fig. 1B). Immunoprecipitation experiments carried out with anti-GFP antibodies revealed that the double mutant completely lost the ability to coprecipitate 14-3-3ɛ, whereas traces of binding could still be observed in some experiments with the single mutants (data not shown). In addition to examining the role of Ser-303 and Ser-307, we also investigated another serine residue (Ser-363) that can be phosphorylated by ERK1 that is localized within another proline-rich region of HSF1, which we have shown previously inhibits HSF1 function when phosphorylated (12, 13). We found, however, that substitution of serine 363 with alanine or glycine did not block HSF1 repression by 14-3-3ɛ and that these mutants behaved essentially like wild-type HSF1, being strongly repressed by the expression of 14-3-3ɛ (Fig. 3B). Phosphoserine 363-mediated repression of HSF1 evidently involves a different mechanism.

FIG. 3.

Effects of 14-3-3ɛ overexpression on the transcriptional activation of the HSP70B promoter by wild-type HSF1 or serine mutants. (A) Schematic representation of the functional domain organization of HSF1. Graphic representations are as indicated: DBD, DNA-binding domain; OLIGO, leucine zipper motif oligomerization domain; RED, regulatory domain; CTA, C-terminal transcriptional activation domain. (B) HeLa cells were cotransfected in triplicate with the HSP70B promoter luciferase reporter plasmid pGLHSP70B and the expression plasmids pHSF1wt, pS303G, pS307G, pS303G-S307G, and pS363A, with (▪) or without (□) pHA-14-3-3ɛ as indicated. In addition, pCMV-lacZ was cotransfected into each culture as an internal control for the transfection efficiency. After the experiment, cells were quenched, and proteins were extracted and assayed for luciferase and β-galactosidase as described in Materials and Methods. Luciferase activity was then normalized to β-galactosidase activity. The relative luciferase activity was then expressed as a percentage of the activity in cells cotransfected with pGLHSP70B and HSF1wt (second column). The relative luciferase values shown in the histogram are the means ± the SDs from three independent experiments. Experiments were carried out four times with reproducible findings each time.

Overexpression of 14-3-3ɛ induces the cytoplasmic localization of HSF1.

Since 14-3-3 has been shown to promote the intracellular translocation of some of its ligands, we next investigated whether such processes were involved in HSF1 repression by 14-3-3ɛ (20). An HSF1-GFP fusion construct (HSF1wt-GFP) was therefore prepared for use in microscopic study of the intracellular localization of HSF1. In control experiments, the GFP fusion did not affect the transcriptional function of HSF1, as indicated by the fact that transfection of HeLa cells with HSF1wt-GFP activated the HSP70B promoter as effectively as wild-type HSF1 (X. Wang and S. K. Calderwood, unpublished results). HSF1wt-GFP was next transfected without or with 14-3-3ɛ into HeLa cells, and the intracellular distribution of HSF1 was examined by using fluorescence microscopy. Our experiments indicate that HSF1wt-GFP was distributed between the nuclear, nuclear-cytoplasmic, and strictly cytoplasmic locations in HeLa cells, as shown previously (36) (Fig. 4A, rows 1 and 2). In contrast, the expression of increasing amounts of 14-3-3ɛ led to a progressive increase in the percentage of cells that expressed HSF1wt-GFP exclusively in the cytoplasm (Fig. 4B). This finding of increased cytoplasmic localization of HSF1wt-GFP in cells overexpressing 14-3-3ɛ is consistent with our earlier finding of repression of HSF1 transcriptional activity by 14-3-3ɛ (Fig. 1) and suggests that inhibition of the transcriptional activity of HSF1 by 14-3-3ɛ may involve nuclear exclusion. In >90% of cells expressing HSF1wt-GFP, the fluorescent signal was predominantly distributed in the cytoplasm when the amount of 14-3-3ɛ plasmid transfected into cells was increased to fivefold of HSF1wt-GFP compared to cells transfected with HSF1wt-GFP alone (Fig. 4B). These experiments provide further evidence implicating 14-3-3ɛ in the nuclear exclusion of HSF1. These conclusions were further supported by experiments using biochemical extraction, which indicated HSF1 exclusion from the nucleus in cells overexpressing 14-3-3ɛ (Fig. 4C). After overexpression, HA-14-3-3 could be detected in the nuclear extract by immunoblotting (Fig. 4C). However, the majority of the transfected 14-3-3ɛ was localized to the cytoplasm as indicated by the immunofluorescence studies (Fig. 4A). Using immunofluorescence microscopy with anti-HA and anti-14-3-3ɛ antibodies, we confirmed that overexpressed HA-14-3-3ɛ was distributed in a similar pattern to the endogenous intracellular 14-3-3ɛ, which is found mainly in the cytoplasm of HeLa cells (Fig. 4A, rows 3 and 4). We also found that endogenous 14-3-3ɛ was coimmunoprecipitated with HSF1wt-GFP when cell extracts were probed with anti-GFP antibodies but not with the double-mutant S303G-S307G-GFP (data not shown).

FIG.4.

Intracellular expression of 14-3-3ɛ induces the cytoplasmic localization of HSF1. (A) HeLa cells were transfected with pHSF1wt-GFP without (rows 1, 2, and 3) or with pHA-14-3-3ɛ (row 4). After 48 h, cells were fixed and incubated with the primary antibodies anti-14-3-3ɛ or anti-HA antibody. Primary antibody staining was detected with a Texas red-conjugated secondary antibody to localize endogenous 14-3-3ɛ (row 3) or coexpressed HA-14-3-3ɛ (row 4) as indicated. HSF1wt-GFP and 14-3-3ɛ were visualized by green autofluorescence or red immunofluorescence as indicated. Cell nuclei were visualized with DAPI autofluorescence. “Merge (1)” indicates colocalization of green or red fluorescence and the DAPI-stained nuclei; “Merge (2)” indicates colocalization of HSF1wt-GFP and HA-14-3-3ɛ. Whole-cell morphology was visualized in the phase-contrast images as indicated. Magnification, ×400. The experiment was carried out three times with reproducible findings each time. (B) Quantitative analysis of the cytoplasmic localization of HSF1wt-GFP in HeLa cells coexpressing increasing amounts of HA-14-3-3ɛ as determined by fluorescence microscopy. The subcellular distribution of HSF1wt-GFP was scored according to whether it was predominantly expressed in the cytoplasm (C bars [▪]). The experiment was carried out three times with reproducible findings each time. (C) Coexpression of 14-3-3ɛ inhibits the nuclear accumulation of HSF1 at 37°C, as determined by biochemical extraction and Western blot analysis. Portions (50 μg) of nuclear extracts (NE) from HeLa cells transfected with pcDNA3.1 HSF1wt alone (lane 1) or with pHA-14-3-3ɛ (lane 2) at 37°C were immunoblotted by using anti-HSF1 antibody to detect the level of HSF1 (upper panel). Quantitation of HSF1 expression (lanes 1 and 2) is shown below the panels. The blot was reprobed with anti-HA antibody (middle panel) to confirm the expression of HA-14-3-3ɛ. Equal protein loading was confirmed by reprobing the same blot with anti-α-actin antibody (bottom panel). The experiment was carried out twice, with similar findings each time. (D) HeLa cells were transfected with pHA-14-3-3ɛ or pHA-P16. After 48 h, cells were fixed and incubated with the primary antibodies anti-HSF1, and anti-HA antibody. Anti-HSF1 was detected with a Texas red-conjugated secondary antibody to localize endogenous HSF1 (rows 1 and 2) and anti-HA was detected with an fluorescein isothiocyanate-conjugated secondary antibody to localize overexpressed HA-14-3-3ɛ (row 3) or HA-P16 (row 4). Endogenous HSF1 and HA-14-3-3ɛ and HA-P16 were visualized by red and green immunofluorescence as indicated. Cell nuclei were stained with DAPI. “Merge (1)” indicates the degree of colocalization of red and green fluorescence of endogenous HSF1 and HA-14-3-3ɛ or HA-P16; “Merge” and “Merge (2)” indicate relative colocalizations of red fluorescence of endogenous HSF1 and the DAPI-stained nuclei. Likewise, to study the effect of 14-3-3ɛ overexpression on intracellular localization of HSF2, cells were fixed and incubated with the primary antibodies anti-HSF2 and anti-HA 48 h after transfection with or without pHA-14-3-3ɛ. Detection with secondary antibodies was carried out as described above. Magnification, ×400. The experiment was carried out three times with reproducible findings. (E) LMB inhibits 14-3-3ɛ-mediated nuclear export of HSF1. HeLa cells cotransfected with pHSF1wt-GFP and pHA-14-3-3ɛ were treated without (Control) or with 10 ng of LMB/ml for 4, 6, and 16 h prior to fixation. Quantitative analysis of the distribution of HSF1wt-GFP in HeLa control cells or in cells coexpressing HA-14-3-3ɛ was performed by using 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 expressed in the cytoplasm (C). The results are the means ± the SDs from three separate experiments.

Since the experiments described above were carried out by using HSF1-GFP chimeric proteins overexpressed in cells, we next examined the behavior of endogenous HSF1. Endogenous HSF1 detected by immunofluorescence with anti-HSF1 antibodies was, as with the HSF1-GFP chimeric protein, distributed between the cytoplasm and the nucleus (Fig. 4D). Cotransfection with HA-14-3-3ɛ led to the relocation of endogenous HSF1 to the cytoplasmic compartment, a finding again similar to the results observed with HSF1-GFP (Fig. 4D). For a further control, we examined the effect of overexpressing another HA-tagged protein, HA-P16, on the intracellular location of HSF1 (Fig. 4D). The expression of HA-P16 did not alter the distribution of HSF1 from its nuclear-cytoplasmic location (Fig. 4D). As a final control, we examined the effect of HA-14-3-3ɛ overexpression on another intracellular protein. For this purpose, we chose to examine heat shock factor 2 (HSF2), a member of the hsf gene family with homology to HSF1 but lacking the regulatory domain that contains Ser-303 and Ser-307. Endogenous HSF2 was detected by immunofluorescence with anti-HSF2 antibodies. HSF2 was distributed evenly throughout the cells, and we detected no evidence of cytoplasmic sequestration on HA-14-3-3ɛ overexpression (Fig. 4D). Indeed, at the higher HA-14-3-3ɛ concentrations, moderately increased levels of HSF2 were detected in the nucleus (Fig. 4D).

LMB inhibits 14-3-3ɛ-mediated nuclear exclusion of HSF1.

14-3-3 has been shown to cause cytoplasmic localization of some target proteins by directing them toward the CRM1-mediated nuclear export pathway (7, 28). We therefore investigated whether active 14-3-3ɛ-mediated nuclear export of HSF1 contributes to its cytoplasmic localization and whether CRM1 is the shuttling receptor responsible for such HSF1 nuclear export. HSF1wt-GFP and HA-14-3-3ɛ were transiently cotransfected into HeLa cells, and the subcellular localization of HSF1wt-GFP was examined by fluorescence microscopy both in the absence and in the presence of the specific CRM1 inhibitor, LMB. LMB binds the nuclear export receptor CRM1 specifically and inhibits its export activity (32, 33). As shown in Fig. 4E, HSF1wt-GFP, when coexpressed with HA-14-3-3ɛ, localized predominantly to the cytoplasm in the absence of LMB treatment (control). However, nuclear accumulation of HSF1wt-GFP was enhanced in the cells treated with LMB (10 ng/ml), and this effect occurred progressively through 4, 6, and 16 h after LMB treatment, with cytoplasmic HSF1 reduced to minimal levels by 16 h (Fig. 4E). These experiments therefore indicate that LMB inhibits 14-3-3ɛ-mediated nuclear export of HSF1 and that CRM1-dependent nuclear export plays a major role in 14-3-3ɛ-mediated HSF1 cytoplasmic retention.

Mutations of ERK1 and GSK3 phosphorylation sites on HSF1 block the 14-3-3ɛ-mediated nuclear exclusion of HSF1.

The experiments shown above indicate that mutation of serine residues in HSF1 to glycine (S303G, S307G, and S303G-S307G) relieve the repression of HSF1 by 14-3-3ɛ (Fig. 1B and 3). We next investigated, therefore, whether serines 307 and 303 are required for 14-3-3ɛ-mediated nuclear export of HSF1. For this purpose, we examined the effect of coexpression of HA-14-3-3ɛ with HSF1wt-GFP and the serine-to-glycine mutants S303G-GFP, S307G-GFP, and S303G-S307G-GFP in HeLa cells. In cells expressing HSF1wt-GFP, ∼61% showed green fluorescence localized to the nucleus. However, the S303G-GFP, S307G-GFP, and S303G-S307G-GFP mutants localized almost exclusively in the nucleus compared to HSF1wt-GFP (Fig. 5A) , further suggesting the requirement for serines 303 and 307 in the cytoplasmic localization of HSF1. After cotransfection with HA-14-3-3ɛ, these mutants retained their predominantly nuclear localization compared to wild-type cells in which cotransfection with HA-14-3-3ɛ caused nuclear exclusion (Fig. 5A, rows 4 to 6, and 5C). Overexpression of HA-14-3-3ɛ significantly increased the cytoplasmic localization of HSF1wt-GFP, and ∼85% of cells expressed HSF1wt-GFP in the cytoplasm, while cotransfection with HA-14-3-3ɛ had little effect on the predominant localization of HSF1 mutants in the nucleus (Fig. 5A to C). These experiments suggest that the phosphorylation of HSF1 on serines 303 and 307 is essential for HA-14-3-3ɛ-mediated nuclear exclusion. In the cells coexpressing the HSF1 mutants and HA-14-3-3ɛ, ∼75% of cells expressing S303G-GFP showed the green fluorescent signal distributed in the nucleus, whereas ∼84% of cells expressing S307G-GFP showed signal in the nucleus and ∼87% of cells expressing S303G-S307G-GFP demonstrated the fluorescence in the nucleus (Fig. 5C). Consistent with the trends in experiments shown in Fig. 1B and C and 3, we found that with HA-14-3-3ɛ overexpression the percentages of cells expressing S307G-GFP and S303G-307G-GFP in the nucleus were similar and higher than those of cells expressing S303G-GFP in the nucleus. This indicates that null mutation of serine 307 is more effective than mutation of Ser-303 in blocking HA-14-3-3ɛ-mediated nuclear exclusion and repression of HSF1.

FIG. 5.

Mutations in the ERK1 and GSK3 phosphorylation sites (Ser-307 and Ser-303) on HSF1 block 14-3-3ɛ-mediated nuclear exclusion of HSF1. (A) HeLa cells expressing the HSF1-GFP serine mutant proteins as indicated without (rows 1 to 3) or with HA-14-3-3ɛ coexpression (rows 4 to 6) were stained with an anti-HA antibody to localize cotransfected HA-14-3-3ɛ and with DAPI to visualize the nuclei. Fixed cells were then analyzed for green autofluorescence (HSF1), red immunofluorescence (14-3-3ɛ), or blue fluorescence (nuclear stain) as described in Fig. 4A. The location of the cells was visualized by phase-contrast microscopy. Magnification, ×400. (B) Relative equilibration of HSF1wt-GFP and the serine mutants between nuclear (N), nuclear-cytoplasmic (N&C), and strictly cytoplasmic (C) distributions in HeLa cells was quantitated as described in Fig. 4D. Microscopic analysis was carried out as in panel A. The values shown in the histograms are the means ± the SDs from three separate experiments. (C) Effect of 14-3-3ɛ expression on relative localization of HSF1wt-GFP and the HSF1 serine-glycine mutants in HeLa cells. The experiment was performed as described in Fig. 4E. Values shown in the histograms are the means ± the SDs from three separate experiments. Columns: N, nuclear; N&C, nuclear-cytoplasmic; C, cytoplasmic.

Overexpression of HA-14-3-3ɛ inhibits HSF1-HSE binding activity.

One likely hypothesis for HSF1 repression by 14-3-3ɛ is that increased nuclear export depletes the nuclear concentration of HSF1 and thus inhibits interaction of HSF1 with heat shock gene promoters. To examine this question, nuclear extracts from HeLa cells transfected with pcDNA3.1 HSF1wt without or with HA-14-3-3ɛ cotransfection were analyzed for HSF1-HSE binding activity. Nuclear extracts were probed with a ³²P-labeled consensus HSE from the HSP70B promoter by using the nondenaturing gel EMSA. We found that HSF1 expressed in cells transfected with HSF1wt alone (Fig. 6 lane 5) or cells cotransfected with HSF1wt and HA-14-3-3ɛ (Fig. 6, lane 7) formed complexes with electrophoretic mobilities similar to those of HSF1-HSE complexes from heat-shocked HeLa cells (Fig. 6, lane 2). These HSF1-HSE complexes were supershifted with anti-HSF1 antibody (Fig. 6, lanes 3, 6, and 8). However, HSF1-HSE binding activity was markedly decreased in cells cotransfected with HA-14-3-3ɛ (Fig. 6, lanes 7 and 8) compared to cells transfected with HSF1 alone (Fig. 6, lanes 5 and 6). Thus, overexpression of HA-14-3-3ɛ reduces the levels of HSF1-HSE binding activity detectable in nuclear extracts. We could not detect the supershifting of HSF1-HSE complexes when HSF1wt was cotransfected with HA-14-3-3ɛ and nuclear extracts were incubated with anti-HA antibody to investigate the presence of 14-3-3 in the HSF1-HSE complexes (Fig. 6, lane 9). This suggests that 14-3-3ɛ is not present in the HSF1-HSE complexes, as would be predicted from the previous results, indicating that 14-3-3ɛ mediates nuclear exclusion of HSF1 (Fig. 4). Therefore, 14-3-3ɛ-mediated inhibition of HSF1 binding to HSE may be due, at least partially, to reduced nuclear accumulation of HSF1, and this may underlie the repression of HSF1 transcriptional activity by HA-14-3-3ɛ demonstrated above.

FIG. 6.

Overexpression of 14-3-3ɛ inhibits HSF1-HSE binding activity. HeLa cells were transfected with empty expression vector pcDNA3.1 (lane 4), with pcDNA3.1 HSF1 alone (lanes 5 and 6), or with pcDNA3.1 HSF1 and pHA-14-3-3ɛ (lanes 7, 8, and 9). Nuclear extracts were then prepared after 24 h of transfection, and EMSA was carried out by using a ³²P-labeled consensus HSE from the HSP70B promoter. Gel supershift assay with anti-HSF1 antibody was carried out as described earlier (52) (lanes 3, 6, and 8). No nuclear extract was added to the incubation in control lane 1. Nuclear extracts from heat-shocked (HS) HeLa cells (lanes 2 and 3) were used as positive controls to demonstrate the migration of HSF-HSE complexes. An arrow indicates the position of HSF1-HSE complex. ³²P-labeled OCT-1 was used as a control to confirm the loading of equal amounts of nuclear extracts (4 μg) in lanes 2 to 9 in the incubations. Quantitation of HSF1-HSE binding is shown below the panels. The experiment was carried out twice times with similar findings.

DISCUSSION

These studies indicate for the first time that HSF1 interacts with the 14-3-3 cell signaling protein family and suggest novel pathways for transcriptional control. Our experiments indicate how the regulatory domain of HSF1 responds to extracellular signals emanating from cell surface binding of external factors, through activation of the ERK cascade, phosphorylation on two conserved serine residues, and association with 14-3-3ɛ. 14-3-3ɛ association then leads to the repression of HSF1 through a mechanism involving cytoplasmic sequestration away from the promoters of hsp molecular chaperone genes.

Our present studies suggest a mechanism whereby HSF1 phosphorylation on serines 303 and 307 regulates interaction with the 14-3-3 family of phosphoserine-binding proteins. It was apparent that although both Ser-307 and Ser-303 contribute to the association with 14-3-3ɛ, phosphorylation on Ser-307 is the more influential modification, although significant levels of 14-3-3ɛ binding are observed with phospho-S303 (Fig. 2). However, phosphorylation of Ser-303 appears to be essential for effective HSF1-14-3-3ɛ interaction in vivo, and mutation of this site blocks 14-3-3ɛ-mediated repression and nuclear exclusion of HSF1 (Fig. 3 and 5). In addition, intracellular overexpression of GSK3, which targets Ser-303, represses HSF1 function and mutation at this site alleviates GSK3-mediated repression (13, 23). Consensus binding motifs in 14-3-3 ligands have been derived by analysis of phosphoserine-oriented peptide libraries. Two consensus sequences (RSXpSXP and RXY/FXpSXP) were derived from these studies (20, 57, 58). Clearly, the HSF1 sequence in the proline rich region of HSF1 (PPpSPPQpSPRV) diverges markedly from this sequence, although the domain containing serines 303 and 307 potentially has features similar to those of the 14-3-3ɛ-binding domain of IGF-I receptor that contains a similar abundance of proline residues (15). It is clear, however, that the HSF1 phosphopeptide binds effectively to 14-3-3ɛ in in vitro binding assays and that mutation of the serines within this region blocks HSF1-14-3-3ɛ interactions in vivo (Fig. 2, 3, and 5). If the HSF1 sequence in the proline rich region is read in inverse orientation, the sequence surrounding Ser-307 (RPpSQPPpS) is closer to the consensus, although it is not clear whether such an inversion is permitted by the stereochemistry of the 14-3-3 phosphoserine-binding cleft. In addition, a phosphopeptide shown previously to bind 14-3-3 (RSPpSPRVC), derived from the cdc25c sequence, competes with the HSF1 phosphopeptide for ³⁵S-labeled 14-3-3ɛ binding, suggesting binding of both phosphopeptides to a similar site in the 14-3-3 protein (41) (Fig. 2E). Our experiments indicate that phosphorylation at both Ser-303 and Ser-307 enhances 14-3-3ɛ binding (Fig. 2C, D, E, and F). Many 14-3-3-binding proteins contain more than one phosphoserine residue, often due to the presence of tandem consensus 14-3-3-binding domains (58). It has been suggested that such tandem domains may contact each member of the 14-3-3 dimer, thus increasing binding affinity (20, 58).

When we tested the in vivo role of serine phosphorylation at Ser-303 and Ser-307 in HSF1 repression by 14-3-3ɛ by a genetic approach, we found that the results were largely predicted by our phosphopeptide-binding studies. Both pSer-303 and pSer-307 were evidently required for HSF1 repression by HSF1, and null mutation of Ser-307 had the stronger inhibitory effect on the blockade of HSF1-mediated transcription by 14-3-3ɛ (Fig. 3).

A compelling hypothesis for the regulation of HSF1 by 14-3-3ɛ is that repression might be mediated by nuclear exclusion, as with a number of other molecules with nuclear functions (7, 16, 20, 41, 45). Indeed, our studies show that 14-3-3ɛ overexpression causes the sequestration of HSF1 in the cytoplasm and, as with repression of the HSP70B promoter, cytoplasmic accumulation is 14-3-3ɛ dose dependent and requires intact serines 303 and 307 (Fig. 1 and 4). The bivalent nature of the 14-3-3 dimer may permit it to function as a molecular scaffold protein, bridging at least two binding proteins (20). Our studies suggest that 14-3-3ɛ association influences the interaction of HSF1 with the nuclear export protein CRM1 and leads to enhanced nuclear export (Fig. 4D), as shown previously for other proteins (7, 28). HSF1 has been shown to contain at least two canonical nuclear localization signals located at either extreme of the leucine zipper sequence in the N terminus (47) and a number of consensus nuclear export signals that have so far not been characterized in molecular studies (Wang and Calderwood, unpublished). 14-3-3ɛ-mediated cytoplasmic sequestration of HSF1 appears to involve CRM1-mediated nuclear export, although the involvement of the other mechanisms cannot be excluded (Fig. 4). Phosphorylation of HSF1 on Ser-303 and Ser-307 apparently shifts the balance between nuclear import and nuclear export toward cytoplasmic localization through the recruitment of 14-3-3ɛ and engagement of the nuclear export machinery (Fig. 2, 3, 4, and 5). Mutation of either serine 303 or serine 307 inhibits nuclear exclusion and results in an almost exclusively nuclear HSF1, presumably by favoring nuclear import, and by conferring resistance to variation in 14-3-3 levels (Fig. 5). Finally, our experiments suggest that nuclear exclusion of phospho-S303-S307-HSF1, through engagement of 14-3-3ɛ or CRM1, may exclude HSF1 availability to Hsp promoters and thus inhibit Hsp synthesis (Fig. 6). It should, however, be noted that in unperturbed HeLa cells, in which 60% of their HSF1 is in the nucleus, only traces of HSF1-HSE binding can be observed (6). Most of such nuclear HSF1 is, however, likely to be in the inactive monomeric form (44). 14-3-3ɛ expression may function to exclude the active, DNA-binding HSF1 trimers from the nucleus as seems to be indicated (Fig. 6). In addition, Cahill et al. (8) have demonstrated an alternative mechanism in the 14-3-3 regulation of daf-16 that could be relevant to the present studies. These authors showed that direct 14-3-3 interaction with daf-16 blocks daf-16-DNA binding. Such a mechanism could be involved with HSF1 exclusion from DNA binding, although it seems unlikely since the DNA-binding domain of HSF1 is remote from Ser-303 and Ser-307 (43).

Regulation of HSF1 by 14-3-3 binding and cytoplasmic localization is overlaid on top of other mechanisms of constitutive repression. HSF1 is under constitutive repression both through intramolecular binding reactions that mask the trimerization domain and the transcriptional activation domains and by association with other cellular proteins such as Hsp90 chaperone complexes (44, 59). Regulation of HSF1 activity by cellular complexes containing molecular chaperones and cochaperones appears to be a crucial level of control (1, 4, 59). Our preliminary studies with the Hsp90-specific inhibitor geldanomycin suggest that Hsp90 does not play a major role in 14-3-3-mediated nuclear exclusion of HSF1, and inhibition of Hsp90 does not override the repressive effects of 14-3-3 expression on HSF1 (results not shown). In addition, we did not observe direct Hsp90 binding to 14-3-3 in immunoprecipitation experiments under conditions in which known Hsp90 binding partners, such as cdc37, were coprecipitated (Grammatikakis and Calderwood, unpublished). These findings suggest that control of HSF1 activity by the Ras/ERK/GSK3/14-3-3 pathway is largely independent of the Hsp90 chaperone machine mechanism of repression and suggest a new, independent pathway of regulation.

Effective repression of HSF1 appears to be essential for normal metabolism. The findings that HSF1 is repressed downstream of the Ras/ERK pathway suggest that HSF1 inhibition may be required for entry into mitosis (12, 13, 23, 29, 30, 37). One possible role that it might play could be in overriding the repressive effects of HSF1 on activation of immediate-early gene promoters, such as c-fos, that are essential for the early events of mitogenesis (10, 24). These effects might be related to the arrest of cells in G₁ observed in cells after heat shock or HSF1 overexpression (6, 27). HSF1 represses Ras-mediated c-fos activation, and such repression might be overcome by the nuclear exclusion of HSF1 via the Ras/ERK/GSK3/14-3-3 pathway outlined above (10). In addition, elevated expression of Hsp chaperones downstream of HSF has been shown to be deleterious for cell proliferation (19, 31). Therefore, inhibition of chaperone expression through the repression of HSF1 might also contribute to mitogenic stimulation. Finally, elevated Hsp expression has been shown to be an important step in malignant transformation, and thus inhibition of HSF1 activity might be important in the suppression of tumorigenesis (51). The repression of HSF1 by ERK phosphorylation and 14-3-3 association may thus play a key role in the maintenance of normal proliferation.

In conclusion, we have shown that HSF1 is repressed by binding to 14-3-3ɛ through a multistep pathway involving the protein kinases ERK1 and GSK3, recruitment of 14-3-3ɛ, cytoplasmic sequestration, and loss of availability to bind to the promoters of heat shock genes. This pathway targets its effects to the regulatory domain of HSF1 and may function to repress the transcription of molecular chaperones during growth and development.