Phosphorylation of HSF1 by MAPK-Activated Protein Kinase 2 on Serine 121, Inhibits Transcriptional Activity and Promotes HSP90 Binding*

Heat shock transcription factor 1 (HSF1) monitors the structural integrity of intracellular proteins and its regulation is essential for the health and longevity of eukaryotic organisms. HSF1 also plays a role in the acute inflammatory response in the negative regulation of cytokine gene transcription. Here we show, for the first time, that HSF1 is regulated by the proinflammatory protein kinase MAPKAP kinase 2 (MK2). We have shown that MK2 directly phosphorylates HSF1 and inhibits activity by decreasing its ability to bind the heat shock elements (HSE) found in the promoters of target genes encoding the HSP molecular chaperones and cytokine genes. We show that activation of HSF1 to bind HSE in hsp promoters is inhibited through the phosphorylation of a specific residue, serine 121 by MK2. A potential mechanism for MK2-induced HSF1 inactivation is suggested by the findings that phosphorylation of serine 121 enhances HSF1 binding to HSP90, a major repressor of HSF1. Dephosphorylation of serine 121 in cells exposed to non-steroidal anti-inflammatory drugs leads to HSP90 dissociation from HSF1, which then forms active DNA binding trimers. These experiments indicate a novel mechanism for the regulation of HSF1 by proinflammatory signaling and may permit HSF1 to respond rapidly to extracellular events, permitting optimal physiological regulation.

Heat shock transcription factor 1 (HSF1) monitors the structural integrity of intracellular proteins and its regulation is essential for the health and longevity of eukaryotic organisms. HSF1 also plays a role in the acute inflammatory response in the negative regulation of cytokine gene transcription. Here we show, for the first time, that HSF1 is regulated by the proinflammatory protein kinase MAPKAP kinase 2 (MK2). We have shown that MK2 directly phosphorylates HSF1 and inhibits activity by decreasing its ability to bind the heat shock elements (HSE) found in the promoters of target genes encoding the HSP molecular chaperones and cytokine genes. We show that activation of HSF1 to bind HSE in hsp promoters is inhibited through the phosphorylation of a specific residue, serine 121 by MK2. A potential mechanism for MK2-induced HSF1 inactivation is suggested by the findings that phosphorylation of serine 121 enhances HSF1 binding to HSP90, a major repressor of HSF1. Dephosphorylation of serine 121 in cells exposed to non-steroidal anti-inflammatory drugs leads to HSP90 dissociation from HSF1, which then forms active DNA binding trimers. These experiments indicate a novel mechanism for the regulation of HSF1 by proinflammatory signaling and may permit HSF1 to respond rapidly to extracellular events, permitting optimal physiological regulation.
Heat shock factor 1 (HSF1) 3 is the transcriptional activator of HSP molecular chaperone genes during stress (1,2). The hsf1 gene plays an essential role in protection of cells from heat shock by regulating the induction of cytoprotective HSP and in protection against the effects of endotoxins through its ability to repress the transcription of proinflammatory cytokines through inhibition of factors involved in cytokine transcription such as NF-IL6 and NFB (3,4). Aging is associated with the degeneration of HSP expression with time and the loss of resistance to cellular oxidants; elevated HSF1 leads to significant increase in lifespan in Caenorhabditis elegans (5)(6)(7). In cancer, the converse situation applies, and malignant transformation is associated with aberrantly high levels of HSP (8,9). These clinical phenomena reflect the role of HSP molecular chaperones in cellular regulation, as either up-or downregulation of HSP expression can profoundly modulate multiple key proteins within the cell (10). It is therefore clear that elucidating the molecular mechanisms that control HSP expression in mammalian cells is essential.
Under normal conditions, most HSF1 is inactive and maintained in a compacted monomeric form (11)(12)(13). Inactive HSF1 lacks the ability to bind to the heat shock elements (HSE) in hsp promoters, is unable to trans-activate hsp genes and fails to repress the promoters of proinflammatory cytokines (13)(14)(15)(16). Activation of HSF1 is a complex process involving monomer to trimer transition and DNA binding; hyperphosphorylation, and capacity to activate target promoters (12,(17)(18)(19). Trimerization of HSF1 is governed by leucine zipper domains in the amino terminus and is subject to intramolecular negative regulation by a fourth leucine zipper in the carboxyl terminus (15). This form of regulation is sufficient to regulate monomer to trimer transition in purified HSF1 in vitro (1). However, additional levels of regulation exist in the cell. Activation is controlled by HSF1 binding to the molecular chaperone HSP90, and HSP90 binding maintains the monomeric state (13,20). This form of regulation is also found in other transcription factor families including the nuclear receptors (21). Activation of HSF1 by heat shock causes the release of HSF1 from HSP90 in a process thought to involve the sequestration of HSP90 by denatured proteins (13). The monomer to trimer transition can also be induced by high concentrations of the non-steroidal anti-inflammatory drugs (NSAIDs) including aspirin, ibuprofen, and sodium salicylate (NaSal) (22,23). These compounds derive the majority of their anti-inflammatory activity through inhibition of the cyclooxygenase enzymes (24,25). However, the induction of trimeric HSF1 involves NSAID concentrations 1000-fold greater than are required for inhibition of cyclooxygenase activity (22,23). At these concentrations, the NSAIDs are effective inhibitors of protein kinases such as RSK2, ERK, and IKK␣ (26,27). This led us to propose the hypothesis that HSF1 activation is negatively regulated by a protein kinase that is inhibited by NSAIDs. HSF1 is subject to reciprocal regulation by proinflammatory and anti-inflammatory signals in myeloid cells; whereas proinflammatory lipopolysaccharides (28) repress HSF1, the NSAIDs activate HSF1 and reverse the effects of lipopolysaccharide (29). Our preliminary studies show that inhibitors of lipopolysaccharide activation of mitogen activated protein kinase-activated protein kinase 2 (MAPKAP kinase 2 or MK2) activate the binding of HSF1 to HSE. Such compounds include tyrosine kinase inhibitors AG126 and herbimycin A, p38 kinase inhibitor SB203580, and NSAIDs sodium salicylate, aspirin, and ibuprofen. 4 MK2 is an inducible kinase activated by p38MAPK and through p38 is regulated by cell stress, lipopolysaccharide, and the cytokines IL1␤ and tumor necrosis factor ␣ (30 -32). MK2 plays an important physiological role in the acute inflammatory response, enhancing the expression of proinflammatory cytokines through mRNA stabilization and increasing expression of inflammatory enzymes including lipoxygenase and phospholipase A (28,(33)(34)(35)(36). Inhibition of HSF1 by MK2, as we suggest here, would be consonant with a role in inflammation, as HSF1 inhibition would relieve the repression on cytokine promoters (3,4,37).
In the current study, we have examined the effects of MK2 on the regulation of HSF1. We showed that activated MK2 is a potent inhibitor of HSF1 and inhibits HSF1 binding to HSE and trans-activation of the HSP70B promoter. Our studies indicate that MK2 directly phosphorylates HSF1 on a specific serine residue 121 (Ser-121) and that this event may mediate some of the intracellular effects of MK2 on HSF1. Mutation of the Ser-121 residue in HSF1 to alanine prevented the inhibitory effects of MK2. Activation of MK2 led to an increase in HSF1-HSP90 binding, whereas the Ser-121 mutation to alanine prevented the effects of MK2 on HSF1-HSP90 binding and permitted HSF1 association with HSE. As HSP90 binding to HSF1 is known to be involved in the inactivation of HSF1, our data suggest a potential mechanism for HSF1 inhibition by MK2. Our experiments therefore indicate that the protein kinase MK2 is an inhibitor of HSF1 and may exert its effects on the protein by direct phosphorylation.
In Vitro Kinase Assays-Purified MK2 (0.1 units, Upstate) was incubated for 20 min at 30°C with 0.25 mM synthetic peptide substrate KKPLNRTLSVASLPG (95% pure) and 0.5 Ci of [␥-32 P]ATP (PerkinElmer Life Sciences) in 25 l of GNM buffer (60 mM ␤-glycerophosphate, 30 mM p-nitrophenyl phosphate, 25 mM MOPS, 15 mM MgCl 2 , 150 mM ATP, 0.1 mM sodium orthovanadate, 5 mM EGTA, and 1 mM dithiothreitol, pH 7.0). Phosphopeptides were isolated on p81 filters (Pierce), washed in ice-cold 75 mM phosphoric acid, and assayed by Cerenkov counting. To detect kinetics of phosphorylation of GST-HSF1 by MK2, fusion protein GST-HSF1 was purified from bacterial lysates as described (4), in vitro kinase assays were performed with 5 g of GST-HSF1 in a 30-l volume of reaction buffer with 0.2 units of MK2 as described above at 30°C for 0, 1, 2, 4, and 6 h, and stopped by addition of SDS sample buffer then analyzed by 10% SDS-PAGE and x-ray film autoradiography.

Mass Spectrometric Matrix-assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) Analysis of Tryptic Peptides-In vitro
phosphorylation of GST-HSF1 by MK2 was carried out for 4 h as described above with radioisotope and then analyzed by 10% SDS-PAGE. The silver-stained GST-HSF1 bands were cut out and in-gel digested with sequencing grade-modified trypsin (Promega), followed by reverse phase HPLC purification and concentration step as described previously (41). The 32 P activity of each fraction was detected using scintillation counting. Molecular masses of 32 P-labeled HPLC fractions from tryptic peptides were analyzed by a MALDI-TOF mass spectrometry (Voyager DESTR).
Cell Culture and Transfection-HeLa cells were maintained in Ham's F-12 (Mediatech) with 10% heat-inactivated fetal bovine serum. Mouse embryonic fibroblasts (mk2ϩ/ϩ) (28) were cultured in Dulbecco's modified Eagle's medium and 10% heat-inactivated fetal bovine serum, 100 g of streptomycin/ml (complete medium). HeLa cells (2.5 ϫ 10 5 cells/well) in 6-well plates were transfected with the plasmids indicated in the figure legends in triplicate using FuGENE 6 (Roche Applied Science) as described (42). pSV-␤-galactosidase plasmid was co-transfected as an internal control for transfection efficiency. pHM6 empty vector was used as a blank plasmid to balance the amount of DNA transfected in transient transfection. Luciferase and ␤-galactosidase activity assays were performed after 24 h of transfection according to the Promega protocol. Luciferase activity was normalized to ␤-galactosidase activity. Results were expressed as relative luciferase activity of the appropriate control.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were incubated with a double-stranded, 32 P-labeled consensus HSE from human HSP70B promoter probe, and analyzed by EMSA as described (38).
Generation of Antibodies Specific for HSF1-Phosphoserine-121-Antibodies were prepared against HSF1-phosphoserine-121 using the phosphopeptide AcRKVT (pS) VSTC-amide (R7C) that spans the Ser-121 site. Phosphopeptide was coupled to bovine serum albumen and injected into rabbits with three subsequent boosting injections. Peptides and antiserum were prepared commercially by BIOSOURCE International. Anti-HSF1-phosphoserine 121 antibody (Ser(P)-121 Ab) was then cleared by passage through an affinity column coupled to unphosphorylated peptide R7C and purified by affinity chromatography on phospho-R7C-agarose.

RESULTS
MK2 Inhibits the Ability of HSF1 to Activate the HSP70B Promoter and Reduces HSF1 Binding to DNA-We first examined the effects of MK2 activation on the ability of intracellular HA-HSF1 to activate the HSP70B promoter. Overexpression of HA-HSF1 led to the activation of heat shock promoter (HSP70B) activity and such HSF1 activation of the HSP70B promoter activity was inhibited by 50 -60% when cells were co-transfected with pMyc-MK2 (Fig. 1A). We next examined potential mechanisms underlying the regulation of HSF1 activity by MK2. Our hypothesis predicts that MK2 may be an inhibitor of the first step in HSF1 activation, which involves the release of HSF1 from inhibitory complexes and the formation of HSE-binding HSF1 trimers (11,12). We tested this hypothesis by examining the effects of Myc-MK2 overexpression on the binding of HSF1 in nuclear extracts to HSEs using EMSA (Fig. 1B). HSF1-HSE binding was assayed in cells transfected with HA-HSF1, as this permitted us to distinguish the transfected HSF1 from endogenous HSF1. Overexpression of HA-HSF1 increased binding to HSE (Fig. 1B, lane 4), and HSF1-HSE complexes could be identified by supershift assay with anti-HA antibody (Fig. 1B, lane 5). Previous studies have shown that expression of HSF1 from exogenous promoters can lead to the trimerization and DNA binding of HSF1 in the absence of stress (12). This may be because of the titration of endogenous inhibitors of trimerization or to the increase in HSF1 concentration facilitating the kinetically rare event of trimer formation. The electrophoretic mobility of the HA-HSF1-HSE complexes derived from the transfected cells was similar to the mobility of such complexes extracted from heatshocked cells (used here as a control) (Fig. 1B, lanes 1 and 2). The rate of migration of the HA-HSF1-HSE complexes in the gels was further  6 and 7). Nuclear extracts were then prepared after 24 h of transfection and analyzed by EMSA. A gel supershift assay with anti-HSF1 and anti-HA antibodies was carried out as described in Ref. 38 (lanes 2, 5, and 7). Nuclear extracts from heat-shocked (HS) HeLa cells (lanes 1 and 2) were used as positive controls to demonstrate the migration of HSF1-HSE complexes. Equal loading of nuclear extracts in lanes 3-7 in the incubations was verified by Western blot analysis with anti-HA antibody. C, HeLa cells were transfected with the indicated plasmids for 24 h then exposed to 20 mM NaSal for 30 min (lanes 2, 4, 6 and 8). Immunoprecipitation (IP) was carried out with anti-HA (Y-11) polyclonal antibody as described under "Materials and Methods." Immunoprecipitated materials were then immunoblotted with anti-phospho-MK2 (Thr-334) polyclonal antibody, anti-Myc monoclonal antibody, or anti-HA monoclonal antibody. Heavy chain IgG was used as a control loading. retarded by anti-HA antibody binding (Fig. 1B, lane 5) migrating at a similar rate to HSF1-HSE complexes from heat-shocked cells bound to anti-HSF1 antibody (Fig. 1B, lane 2). Overexpression of Myc-MK2 strongly inhibited HA-HSF1 binding to HSE whether incubated without or with anti-HA antibody (Fig. 1B, lanes 6 and 7). We also examined the physical association of overexpressed Myc-MK2 with HA-HSF1. Myc-MK2 overexpression led to its activation as indicated by phosphorylation on threonine 334 (data not shown). Such active, overexpressed Myc-MK2 became bound to HA-HSF1 and was recovered in anti-HA immunoprecipitates and detected by immunoblot assay with antibodies specific for the Myc tag (total Myc-MK2) or phospho-MK2 (Thr(P)-334; Thr-334-phosphorylated Myc-MK2) (Fig. 1C). NaSal appeared to reduce Thr(P)-334 levels. The reason for this is not clear, although in similar studies carried out on RSK2, NaSal reduced RSK2 autophosphorylation by unknown mechanisms (27). In control cells in which HA-HSF1 was not transfected, we did not observe phospho-MK2 in the HA immunoprecipitates (Fig. 1C).
MK2 Directly Phosphorylates HSF1-We next examined the hypothesis that HSF1 is a direct substrate for phosphorylation by MK2 and that this phosphorylation may mediate the inhibitory effects of MK2 ( Fig.  2A). When a purified GST-HSF1 fusion protein was incubated with MK2 in vitro, we observed 32 P-incorporation into GST-HSF1 in a timedependent manner that reached a maximum by 4 h. We estimated incorporation at this point at ϳ1.3 mol of orthophosphate/mol of GST-HSF1 ( Fig. 2A). However as the peptide mapping analysis shown later indicates phosphorylation of GST-HSF1 is at at least 6 sites, phosphorylation of individual serines/threonines is likely to be substoichiometric. We next investigated whether this interaction could be inhibited by NaSal. We incubated GST-HSF1 with purified MK2 and determined the effect of incubation with 20 mM NaSal on 32 P incorporation. This concentration of NaSal inhibited 32 P incorporation by ϳ50% (Fig. 2B, compare lane 2 with 3). We next examined in more detail whether MK2 activity is inhibited by the doses of NaSal that activate the binding of HSF1 to the hsp promoters in vivo (22,23) by using a well characterized substrate peptide derived from the hsp27 sequence (32). The experiments indicate that MK2 activity is inhibited in a dose-dependent manner, with ϳ50% inhibition at a concentration of 20 mM, the concentration that inhibits HSF1 phosphorylation (Fig. 2, B and C) and activates the binding of HSF1 to HSE in vivo (22,23).
We next investigated the identity of the amino acid residues within HSF1 phosphorylated by MK2. GST-HSF1 was phosphorylated by MK2 as in Fig. 2A in the presence of [␥-32 P]ATP for 4 h at 30°C in vitro, analyzed by SDS-PAGE, and visualized by silver staining as in Fig. 2D. (GST, unconjugated to HSF1 was not phosphorylated in these circumstances Fig. 2D). The non-phospho-HSF1 (control, Fig. 2D, upper panel,  lane 3) and phospho-HSF1 (Fig. 2D, upper panel, lane 4) bands were excised from the gel and in-gel digested with sequencing grade-modified trypsin, followed by reverse phase HPLC purification and concentration. Radioactivity in each HPLC fraction was determined by liquid scintillation counting, 32 P-containing fractions eluted by HPLC were then analyzed by MALDI-TOF mass spectrometry (Fig. 2, D and E). Two major phosphopeptide species were resolved by this analysis, including one (118 -127) with the sequence (R)KVTSVSTLK(S) containing four potential phosphorylation sites and a second peptide 523-529 (K)AKDPTVS(-) from the extreme carboxyl terminus of HSF1 containing Thr-527 and Ser-529 (Fig. 2, E and F). Only one of these potential phosphate acceptor sites, serine 121 resembles a consensus site for MK2 phosphorylation (HXR/KXXS/T) (Fig. 2F). This motif resembles a MK2 site in possessing an upstream lysine residue in Ϫ3 position and hydrophobic residue at Ϫ6 relative to serine 121. The sequence deviates from the consensus derived in studies using synthetic peptides and purified MK2 carried out by Stokoe et al. (32) in that Ile-115 is at the Ϫ6 position relative to serine 121 rather than Ϫ5 as would be predicted from the consensus (Fig. 2F). Serine 121 is highly conserved across mammalian and avian species, consistent with an important role in HSF1 regulation (Fig. 2G). This sequence was not conserved in yeast or Drosophila HSF1 suggesting that this form of regulation may be confined to vertebrates. Our mass spectrometry analysis carried out in duplicate therefore indicates MK2 phosphorylation of HSF1 in vitro at these sites only. Previous analysis of HSF1 phosphorylation by MK2 by two-dimensional peptide mapping also indicated phosphorylation in peptides corresponding in chromatographic behavior to the peptides indicated in Fig. 2F. 5 Role of Serine 121 in Repression of HSF1 by MK2-We next examined the potential functional roles of all the serine and threonine residues in HSF1 that we identified as MK2 regulated phosphoacceptor sites. We examined their potential role in regulation by point mutations of residues corresponding to these phospho-Ser/Thr residues characterized by MALDI-TOF ( Fig. 2 and Tables 1 and 2). We also constructed HSF1 mutants with amino acid substitutions at multiple sites, as the MALDI-TOF analysis suggested that some of the phosphopeptides isolated after MK2 treatment are multiply phosphorylated ( Table 1). The HSF1 mutant proteins were initially tested for their ability to activate the transcription of a hsp70B-based promoter reporter construct after cotransfection into cells in vivo, as described in Fig. 1A. The HSP70B promoter was strongly inhibited when cells were co-transfected with wild-type HSF1 and Myc-MK2, consistent with the result in Fig. 1A, whereas overexpressed dominant negative MK2 (DN) failed to inhibit HSF1 (Fig. 3A). Transactivation of HSF1 was enhanced by mutation of serine 121 to alanine, and the activity of the S121A construct was not inhibited by MK2 overexpression, suggesting that Ser-121 mediates MK2 inhibition of HSF1 activity (Table 1 and Fig. 3A). Mutation of Ser-121 to aspartate (S121D), in an effort to mimic Ser(P)-121-HSF1, led to an HSF1 molecule with diminished capacity for HSF1 activation (Fig. 3A), consistent with the hypothesis that negative charge at Ser-121 may be inhibitory for HSF1 activity. Mutations at other potential MK2 phosphorylation sites within peptide VTSVSTLK to yield T120A, S123A, and T124A were not effective, and the mutants behaved essentially as wild-type HSF1 (Table 1). Likewise the T527A and S529A mutants behaved like wild-type HSF1 (Table 1). When multiple mutations containing the S121A mutant were prepared, we found that all behaved essentially as the S121A single mutant, and these included T120A/S121A, S121A/S123A, T120A/S121A/S123A, T120A/S121A/ T527A, and T120A/S121A/S529A (Table 1). In addition, the T527A/ S529A double mutant behaved like wild-type HSF1 (Table 1). Thus, although multiply phosphorylated HSF1 peptides can be isolated from HSF1 after in vitro phosphorylation by MK2, these changes appear to have little significance for HSF1 regulation in vivo. Only Ser-121 appears to play a significant role in the response of HSF1 to MK2 in vivo (Table  1). We also screened the mutant forms of HSF1 for ability to bind to HSE in the in vitro EMSA conditions described in Fig. 1B. EMSA was carried out on nuclear extracts from cells after overexpression of wild type HSF1 and mutants in vivo ( Table 2). The results of these experiments paralleled the findings in the trans-activation experiments ( Table 1), indicating that all constructs with the exception of S121A mutants behave in an essentially similar manner to wild-type HSF1 (Tables 1 and  2). The effects of Ser-121 mutation on HSF1-HSE binding are shown in Fig. 3B. Wild-type HA-HSF1 is induced to bind HSE after overexpression in cells in vivo, and its identity is validated by anti-HA supershift 5 B. Chu and S. K. Calderwood, unpublished data.

FIGURE 2. MK 2 directly phosphorylates HSF1 in vitro.
A, GST-HSF1 fusion protein was incubated with purified MK2 at the different time course (0, 1, 2, 4, and 6 h) with [␥-32 P]ATP labeling as described under "Materials and Methods" and then the reaction was stopped by addition of SDS sample buffer before analysis by 10% SDS-PAGE and x-ray film autoradiography. Quantitation of relative density (mol of phosphate/mol of GST-HSF1) from ␥-32 P-labeled GST-HSF1 bands is shown below the panels. B, phosphorylation of GST-HSF1 fusion protein by purified MK2 with and without 20 mM NaSal. GST-HSF1 was incubated with MK2 as described under "Materials and Methods" and then analyzed by 10% SDS-PAGE and x-ray film autoradiography. C, dose-response curve for the inhibition of purified MK2 by NaSal. Purified MK2 was incubated for 20 min with specific peptide substrate (Fig. 3B, lanes 4 and 5). Binding of HSF1 to HSE is enhanced by the mutation HA-S121A, and the intensity of the supershifted band is strongly enhanced (Fig. 3B, lanes 6 and 7). The binding of HSF1 to HSE was partially inhibited by HA-S121D mutation, and the intensity of the supershifted band was markedly reduced (Fig. 3B, lanes 8 and 9). These effects are not as apparent in the non-supershifted complex and this may be because this complex contains both endogenous wild-type HSF1 as well as HA-S121D, whereas the anti-HA antibody supershifted complex only contains HA-S121D free of endogenous HSF1 (Fig. 3B, lanes 8  and 9). Our data indicate a role for serine 121 in the repression of HSF1 binding to HSE and the HSP70 promoter activity by MK2. It is significant that this site has recently been shown to be phosphorylated in HSF1 in human cells in vivo (44). Although those studies do not explore the potential regulatory role of serine 121 in HSF1 activity in vivo, they do concur with the present study that this site is one of the serine residues in HSF1 that can be phosphorylated under physiological conditions.
Phosphorylation of HSF1 on Serine 121 Promotes HSP90 Binding to HSF1-Ability of HSF1 to bind DNA is regulated by reversible binding to HSP90 as well as a number of co-chaperones, and HSP90 complexes are thought to maintain HSF1 in an inactive, non-DNA binding form (13,20). We have therefore examined the hypothesis that phosphorylation of HSF1 by MK2 may influence HSP90 binding and thus repress HSF1. When we examined the distribution of HSF1, HSP90, and MK2 by gel filtration, we found some overlap in their retention by the column. The hydrodynamic behavior of HSF1 from whole cell extracts indicates its association in complexes of median M r 440 -232 kDa (Fig. 4A), consistent with a previous report (45). Although HSP90 fractions tended toward higher M r fractions (median 440 kDa), and MK2 fractions tended to a smaller complex (Ͻ158 kDa) (Fig. 4A), two areas of potential HSP90 and MK2 overlap with HSF1 were observed (Fig. 4A, at fraction 45, 83, lanes 4 and 15), that could correspond to an HSF1-HSP90-MK2 complex.
We next carried out co-immunoprecipitation experiments from cells overexpressing FLAG-HSF1 with or without Myc-MK2 or Myc-MK2 (DN) to analyze the potential interaction of FLAG-HSF1 and endogenous HSP90 (Fig. 4B). Previous studies have indicated that co-immunoprecipitation of HSF1 and HSP90 is difficult to demonstrate evidently because of the transient nature of the interaction (46). However we were able to detect FLAG-HSF1 co-immunoprecipitation with HSP90 under basal conditions, presumably because of the increased levels of overexpressed FLAG-HSF1 (Fig. 4B, lane 2). Overexpression of Myc-MK2 led to increased FLAG-HSF1/HSP90 binding above the endogenous level (Fig. 4B, compare lane 2 with 3, anti-HSP90 immunoblot), but no changes were seen when Myc-MK2 (DN) were overexpressed (Fig. 4B,  lane 4). No FLAG-HSF1/Myc-MK2 co-association was seen if the immunoprecipitation was carried out under denaturing conditions (data not shown) performed as described (38), indicating true co-immunoprecipitation of the molecules under native conditions (Fig. 4B,  lane 2). As NaSal is an MK2 inhibitor (Fig. 2), we next examined its role in HSF1-HSP90 binding (Fig. 4C). We used here HA-HSF1 rather than FLAG-HSF1 to ensure that the results were not influenced by the type of affinity tag used in the experiments. Incubation of cells with 20 mM NaSal led to an ϳ50% decrease in HA-HSF1 binding to endogenous HSP90, consistent with a role for MK2 in enhancing the association of HSF1 with HSP90 (Fig. 4, C and D, compare lane 3 with 4, anti-HSP90 immunoblot). To test a role for serine 121 in these effects, we then examined the effect of S121A mutation on HSP90 binding. HSP90 binding to HSF1 was markedly reduced with S121A mutation (Fig. 4, C and  D, lane 7, anti-HSP90 immunoblot). However the S121D mutant form of HSF1 bound effectively to HSP90, indicating a role for negative charge in the 121 position in HSP90 binding (Fig. 4, C and D, lane 8

TABLE 2
Effects of MK2 overexpression on HSF1-HSE binding activity by wild-type HSF1 and a series of the mutants ND, not determined.
To examine more directly the role of serine 121 in HSF1 activity, we prepared an antibody that specifically reacts with the phosphoserine 121-HSF1. As can be seen, the immunoblot experiments using anti-Ser(P)-121 antibodies indicated that immunoprecipitated HA-HSF1 is constitutively phosphorylated on serine 121, and this activity is inhibited by NaSal and stimulated by ϳ20% through Myc-MK2 overexpression (Fig. 4, E and F, lanes 3-6). In cells transfected with HA-S121A or HA-S121D, little evidence of immunoreactivity is seen with or without the addition of NaSal, indicating the specificity of the antibody for Ser(P)-121 (Fig. 4, E and F, lanes 7-10). The degree of phosphorylation of HSF1 on serine 121 thus correlates well with its ability to bind to HSP90 (Fig. 4). This small increase in HSF1 phosphorylation appears to be sufficient to inhibit the DNA binding and trans-activation of transfected HA-HSF1 (Fig. 1, A and B). It is known that only a small fraction FIGURE 3. Effects of MK2 overexpression on the transcriptional activation of the HSP70B promoter and HSF1-HSE binding activity by wildtype HSF1 or mutants. A, HeLa cells were co-transfected in triplicate with pGLHSP70B and expression plasmids pHA-HSF1 wt, pHA-S121A, pHA-S121D with or without Myc-MK2; Myc-MK2 dominant negative (DN) mutant as indicated. In addition, pCMV-lacZ plasmid was co-transfected into each culture as an internal control for transfection efficiency. Experiments were carried out four times with reproducible findings as described in the legend for Fig. 1. B, mk2ϩ/ϩ cells were transfected with empty vector pHM6 (lane 3), with pHA-HSF1 wt alone (lanes 4 and 5), with pHA-S121A (lanes 6 and 7) or pHA-S121D (lanes 8 and 9). Nuclear extracts were then prepared after 24 h of transfection, and EMSA and gel supershift assay were carried out as described in the legend for Fig.  1. Nuclear extracts from heat-shocked (HS) HeLa cells (lanes 1 and 2) were used as positive controls to demonstrate the migration of HSF1-HSE complexes. Equal loading of nuclear extracts in lanes 3-9 in the incubations was verified by Western blot analysis of the same nuclear extracts with anti-HA antibody. C, quantitation of relative HSF1-HSE binding from B is shown using densitometric analysis of the shifted or supershifted bands on the autoradiographs.  2, 4, and 6). Whole cell extracts were subjected to immunoprecipitation with anti-HA (Y-11) polyclonal antibody and then immunoblotted with anti-HSP90, anti-Myc, and anti-HA antibodies. The heavy chain IgG band was used as a loading control. The experiment was repeated twice with consistent findings. D, relative levels of HSP90 and Myc-MK2 in the HA-HSF1 immunoprecipitate were quantitated by densitometry. E, HeLa cells were transfected with the indicated plasmids. After 24 h, transfected cells were treated with 20 mM NaSal for 30 min. Immunoprecipitation was carried out using anti-HA (Y-11) polyclonal antibody. Immunoprecipitated materials were then analyzed by Western blot using anti-Ser(P)-121 and anti-HA antibodies. Experiments were carried out in duplicate with similar findings. F, quantitation of relative density of phospho-HSF1 (pS121) from A is shown. of transfected HSF1 is able to bind HSE and activate hsp transcription and this fraction appears to be the fraction that is not phosphorylated at serine 121 or bound to HSP90 (12,47). Forced expression of MK2 appears to lead to phosphorylation of free HSF1 on serine 121, and this is associated with HSP90 binding and inhibition of HSE binding (Fig. 4, B and E).
Heat Shock Can Override the Effects of MK2 on HSP90 Binding to HSF1-When cells are exposed to heat shock, HSP90 rapidly dissociates from HSF1, permitting HSF1 to migrate to the nucleus and activate HSP promoters and repress cytokine genes despite the fact that MK2 is strongly activated by the same heat shock conditions (30). We therefore examined whether heat shock can overturn the ability of MK2 to stabilize HSF1-HSP90 interactions. Indeed, we observed rapid dissociation of HSF1 from HSP90 in heat-shocked cells even in the presence of overexpressed Myc-tagged MK2 (Fig. 5A). Exposure to heat shock decreased the association of the Myc-MK2 with HSF1 and led to dephosphorylation of Ser-121 (Fig. 5A). Dephosphorylation of Ser-121 may thus participate in the rapid escape of HSF1 from HSP90-mediated repression observed during heat shock (Fig. 5A).
As many of these experiments have been carried out on HSF1 overexpressed in cells to permit the examination of serine 121 mutants, we also carried out studies on the role of Ser-121 phosphorylation in endogenous HSF1 (Fig. 5B). HeLa cells were either exposed to sodium salicylate or heat shock and then analyzed by immunoprecipitation with anti-Ser(P)-121 antibodies. In control cells endogenous HSF1 was immunoprecipitated by the antibody, whereas NaSal or heat treatment markedly reduced the degree of immunoprecipitation indicating Ser-121 phosphorylation under resting conditions and significant loss of Ser-121 phosphorylation under these circumstances (Fig. 5B).

DISCUSSION
Our experiments suggest a mechanism whereby inflammatory signals can repress HSF1 through downstream convergence on MK2 and inhibition of endogenous hsp expression in monocytes and macrophages. We show that MK2 directly inhibits HSF1 activity through a mechanism involving phosphorylation on serine 121 and that anti-inflammatory drugs can antagonize this effect and activate HSF1 by direct inhibition of MK2 (Figs. 1 and 2). Our experiments suggest that MK2 exerts its effects at least partially by phosphorylation of HSF1 on Ser-121 and by inhibiting association of the phosphorylated HSF1 with HSE in hsp gene promoters (Figs. 1 and 3). MK2 can thus, by inhibiting HSF1, effectively repress endogenous hsp expression. We have also identified a potential mechanism whereby HSF1 phosphorylation on serine 121 by MK2 regulates HSF1. Previous studies have shown that HSF1 is maintained in an inactive monomeric form by its association with inhibitory protein complexes containing HSP90 (13). Our current studies suggest that HSF1 regulation by serine 121 phosphorylation involves this mode of control as HSF1 binding to HSP90 is enhanced by serine 121 phosphorylation or faux phosphorylation through serine 121 to aspartate mutation and decreased when HSF1-HSP90 binding is reduced by NaSal treatment and completely inhibited by serine 121-alanine mutation (Fig. 4). Constitutive repression of HSF1 under resting non-stress conditions has been previously ascribed to the cooperative engagement of two types of regulation, including 1) intramolecular masking of the trimerization domains and 2) binding of the inhibitory HSP90-containing molecular chaperone complex (12,14,15). HSF1 activation by stress was suggested to involve the sequestration of HSP90 in protein aggregates generated by protein damage (13). The addition of a third component, involving serine 121, to these regulatory interactions may offer a sharper, more flexible transcriptional control of HSF1 activity and greater responsiveness to physiological circumstances such as activation of the acute phase response. Previous studies show that the linker region enclosing Ser-121 controls the monomer to trimer transition (48). In addition, serine 121 is immediately adjacent to one of the bipartite nuclear localization signals in HSF1 (116 -118) and HSP90 binding adjacent to serine 121 could act to mask this domain (49). In the acute phase response, cytokine expression is ultimately followed by repression in the resolution of the response (50,51). This may involve the febrile activation of HSF1, which can repress cytokines directly by binding to their promoters or indirectly by the induction of HSP70, an inhibitor of the proinflammatory transcription factor NFB (52). Under these circumstances, MK2 can be overridden by two potential mechanisms. Firstly, accumulation of denatured/aggregated proteins that sequester HSP90 and activate HSF1 independently of Ser-121 phosphorylation (13,20). Secondly, our data show that Ser-121 becomes at least partially dephosphorylated by stress even when MK2 is active (Fig. 5B). However, the exact role of Ser-121 phosphorylation in HSF1 activation by stress is not yet clear, and more information is required to determine the quantitative levels of Ser-121 phosphorylation after stress and whether Ser-121 dephosphorylation (as in Fig. 5) plays a causal role. It is also possible that the effects of MK2 on HSF1 may also involve other cellular effects of MK2, which is a pleiotropic factor involved in a range of cellular pro- FIGURE 5. Effects of heat shock on serine 121 phosphorylation. A, HeLa cells were transfected with the indicated plasmids for 24 h, transfected cells were either untreated or treated with heat shock at 43°C for 30 min. Whole cell extracts were immunoprecipitated (IP) with anti-HA (Y-11) antibody as indicated. Immune complexes were analyzed by SDS-PAGE and immunoblotted with anti-HSP90, anti-Myc, anti-Ser(P)-121, and anti-HA antibodies. The heavy chain IgG band was used as a loading control. Experiments were repeated reproducibly three times. B, phosphorylation of endogenous intracellular HSF1 on serine 121. HeLa cells were grown in Ham's F-12 medium with 10% heat-inactivated fetal bovine serum and heatshocked at 43°C for 30 min or exposed to 20 mM NaSal for 30 min. Immunoprecipitation was carried out with anti-phospho-HSF1 (serine 121). HeLa lysate (lanes 1-3) and immunoprecipitated materials (lanes 4 -6) were then analyzed by Western blot using anti-HSF1 antibody. Experiments were carried out in duplicate with similar findings. The histogram shows relative levels of HSF1 in the anti-phospho HSF1 (serine 121) immunoprecipitate were quantitated by densitometry.
cesses (28,36). Inhibition of HSF1 by inflammatory signals could also involve additional phosphorylation sites in HSF1. Lipopolysaccharide can lead to the activation of the ERK cascade, and ERK activation has been shown to lead to repression of HSF1 through inhibitory phosphorylation site serine 307, which can lead to the nuclear export of HSF1 (38,(52)(53)(54).
HSP90 has been shown to interact with three classes of polypeptides in the cell, including (a) co-chaperones such as p23 and FKBP52, (b) client proteins that bind stably to HSP90, most notably protein kinases and steroid hormone receptors, and (c) unfolded proteins and peptides that are chaperoned by HSP90 (55,56). HSF1 is similar to other HSP90 client proteins in that it forms a supercomplex with HSP90, immunophilins, and p23 that is functionally inactive (13,20). Its interaction, however, differs in quality from many other clients in that dissociation from HSP90 during heat shock or treatment with HSP90 inhibitors does not destabilize HSF1 and does not target it for destruction as with a multitude of other substrates (57). It therefore seems that the HSP90 complex functions as a reversible HSF1 repressor and is not needed for HSF1 stabilization. The physicochemical nature of the interaction may thus be different from other HSP90 clients such as steroid hormone receptors in which the chaperone domains of HSP90 bind a disordered/ unstable region of the receptor (57,58). However, it is also possible that other HSP90-binding proteins may be regulated in a phosphorylationdependent manner like HSF1, as observed recently for the asialoglycoprotein receptor (59). Serine 121 is located in the linker domain of HSF1, between the DNA binding domain and the first coiled-coil region (49).