Activation of Heat Shock Factor by Alkylating Agents Is Triggered by Glutathione Depletion and Oxidation of Protein Thiols*

Transcriptional activation of heat shock protein genes is a common response to proteotoxic stress. Many drugs and chemicals that form reactive electrophiles modify protein structure by binding covalently to nu-cleophilic functional groups. Although many of these agents also activate transcription of the inducible member of the hsp70 gene family, it is not clear if covalent modification of cellular proteins per se is sufficient. Iodoacetamide (IDAM) is a prototypical alkylating toxi- cant that induces hsp70 transcription. However, IDAM-induced cell death is indirectly linked to protein alkylation through depletion of glutathione, induction of oxidative stress, and increased lipid peroxidation. Therefore, we determined if any of these secondary cytotoxic events might lead to activation of hsp70 tran- scription. IDAM treatment increased hsp70 transcription by activating heat shock transcription factor-1 (HSF1). The addition of antioxidants and iron or calcium chelators prevented cell death but did not prevent hsp70 transcription or HSF1 activation. However, the protein synthesis inhibitor cycloheximide blocked acti- vation of hsp70 by low concentrations of IDAM. Further-more, the addition of dithiothreitol (DTT) after IDAM removal blocked hsp70 transcription and HSF1 activation without altering IDAM binding. DTT had no effect on activation of HSF1 by hyperthermia. After IDAM treatment, cellular nonprotein and protein thiols had decreased to less than 20 and 70%, respectively, of the value in control cells. DTT treatment in situ prevented the loss of cellular protein thiols and blocked the formation of high molecular weight protein aggregates. Thus, alkylation of

Transcriptional activation of heat shock protein genes is a common response to proteotoxic stress. Many drugs and chemicals that form reactive electrophiles modify protein structure by binding covalently to nucleophilic functional groups. Although many of these agents also activate transcription of the inducible member of the hsp70 gene family, it is not clear if covalent modification of cellular proteins per se is sufficient. Iodoacetamide (IDAM) is a prototypical alkylating toxicant that induces hsp70 transcription. However, IDAMinduced cell death is indirectly linked to protein alkylation through depletion of glutathione, induction of oxidative stress, and increased lipid peroxidation. Therefore, we determined if any of these secondary cytotoxic events might lead to activation of hsp70 transcription. IDAM treatment increased hsp70 transcription by activating heat shock transcription factor-1 (HSF1). The addition of antioxidants and iron or calcium chelators prevented cell death but did not prevent hsp70 transcription or HSF1 activation. However, the protein synthesis inhibitor cycloheximide blocked activation of hsp70 by low concentrations of IDAM. Furthermore, the addition of dithiothreitol (DTT) after IDAM removal blocked hsp70 transcription and HSF1 activation without altering IDAM binding. DTT had no effect on activation of HSF1 by hyperthermia. After IDAM treatment, cellular nonprotein and protein thiols had decreased to less than 20 and 70%, respectively, of the value in control cells. DTT treatment in situ prevented the loss of cellular protein thiols and blocked the formation of high molecular weight protein aggregates. Thus, alkylation of proteins is insufficient to activate hsp70 transcription and DNA binding of HSF1. However, cellular thiol-disulfide redox status and formation of disulfide linked aggregates of cellular proteins are linked to HSF1 activation and hsp70 transcriptional activation.
The cellular response to stress involves a universally conserved sequence of events, most notably elevated expression of proteins referred to generically as stress proteins (1)(2)(3). Induction of heat shock proteins (HSPs), 1 a family of highly con-served proteins (4,5) is elicited by a variety of stresses including elevated temperature or exposure to amino acid analogs, heavy metals, oxidants, and chemical agents (6). The increased HSPs enhance cell survival by repairing and/or preventing protein damage, a role linked to the ability of HSPs to act as "molecular chaperones" (4,5,7).
Inducers of the heat shock response have in common the ability to denature proteins and promote aggregation (6,8), leading Hightower (9) to propose that proteotoxicity is an important trigger for the heat shock response. Accordingly, microinjecting denatured proteins into Xenopus oocytes is sufficient to elicit the heat shock response (10,11). At the same time, HSPs were proposed to function as molecular chaperones during protein folding and to prevent aggregation of nascent polypeptide chains (12,13). Subsequent studies showed that HSPs prevent aggregation of denatured proteins, resolubilize protein aggregates that have already formed, and assist in refolding denatured proteins (4,5,14,15). The ability of HSPs to retain denatured proteins in soluble form is a critical step in proteolysis of damaged proteins (16 -19). Thus, HSPs serve as a switching point in recognition of reversibly and irreversibly damaged proteins.
A plausible molecular mechanism underlying heat activation of the inducible hsp70 gene, also called hsp72, has been proposed (20,21). In contrast to the constitutive form of hsp70, also called hsc70 or hsp73, inducible hsp70 transcription is up-regulated dramatically by thermal stress. Transcriptional activation is mediated by binding of HSF1, a member of the heat shock factor (HSF) family, to heat shock elements (HSEs) located within the hsp70 gene (22). It has been proposed that HSF1 protein is inactive in unperturbed cells but has heatinducible DNA binding activity. Activation of HSF1 may involve trimerization, phosphorylation, and, in some cells, translocation from the cytoplasm to the nucleus (23)(24)(25)(26)(27)(28). In addition, direct or indirect interactions with HSPs may repress HSF1 DNA binding in unstressed cells (29 -32). According to this hypothesis, increased binding of HSPs to thermally denatured proteins (15) derepresses HSF1, which activates transcription (21).
Chemicals that form reactive intermediates also elicit a heat shock response (6,8). For example, chemotherapeutic agents form electrophiles that covalently bind to cellular macromolecules, including proteins, and increase HSP70 synthesis (33,34). Likewise, reactive metabolites of drugs and pollutants covalently bind to proteins via arylation, alkylation, and acylation (35) and increase HSP synthesis (36). Intuitively, covalent modification alters protein structure; however, it is unclear if this is sufficient to induce hsp70 expression in cells. Because alkylating agents also deplete glutathione (GSH) and induce oxidative stress (37,38), both of which induce a heat shock response (15, 39 -41), it is also possible that activation of hsp genes is triggered by events secondary to covalent modification of proteins. Indeed, our studies with the thioacylating toxicant, S-(1,2-dichlorovinyl)-L-cysteine, indicate that perturbation of the thiol-disulfide redox status plays a role in transcription of hsp70 (36). Because alkylating agents are important as both drugs and pollutants, understanding the how they activate the heat shock response is warranted.
Iodoacetamide (IDAM) is a prototypical cytotoxic alkylating agent that reacts with protein-cysteinyl sulfhydryl groups to form S-acetamido thioether protein adducts. Although adduct formation is associated with HSP synthesis and cell death (42), covalent binding of IDAM does not cause cell death directly. Rather, cell death is linked to depletion of soluble GSH, oxidative stress, and lipid peroxidation (37). Accordingly, IDAM cytotoxicity is inhibited by the thiol reducing agent DTT, the iron chelator desferroxamine (DFAM), and the antioxidant N,NЈ-diphenyl-phenylenediamine (DPPD), none of which affect IDAM binding per se. Thus, IDAM is an excellent model to determine if cytotoxic signals secondary to covalent binding activate hsp70 transcription. Herein we show that IDAM-induced HSF1 DNA binding and hsp70 transcription are blocked by DTT and inhibitors of protein synthesis but not by antioxidants or chelators. DTT also prevents formation of disulfide linked proteins (PSSP) without decreasing IDAM covalent binding, thus clearly segregating protein alkylation from HSF activation. We propose that activation of HSF1 DNA binding and hsp70 transcription by IDAM, and perhaps other alkylating agents, is not due to covalent binding to proteins but is caused by PSSP formation secondary to GSH depletion. Thus, perturbation of the cellular thiol-disulfide redox status may be an important common signal for activation of the heat shock response by toxicants which form reactive intermediates.

EXPERIMENTAL PROCEDURES
Materials-Fetal bovine serum and Dulbecco's modified Eagle's medium (DMEM) were obtained from Life Technologies, Inc. LLC-PK1 cells, a porcine renal epithelial cell line with characteristics of the proximal tubule epithelium (43), were obtained from American Type Culture Collection (Rockville, MD) at passage 195 and were used from passage 205 to 215. DPPD was obtained from Kodak. The acetoxymethyl ester of EGTA was purchased from Molecular Probes (Eugene, OR). All other chemicals were obtained commercially.
Cell Cultures, Toxicant Treatment, and Cytotoxicity Assays-Cell culture and toxicant treatments were carried out basically as described previously (37,38). LLC-PK1 cells were maintained at 37°C in a humidified incubator with an atmosphere of 95% air and 5% carbon dioxide in DMEM supplemented with 10% fetal bovine serum. Medium was added to 6-well plates or 10-cm dishes in a volume of 2.4 ml/well or 13 ml/dish, respectively. Confluent LLC-PK1 cells were treated with IDAM for 15 min in Earle's balanced salt solution (EBSS), then washed twice with phosphate-buffered saline (PBS), and allowed to recover in DMEM containing 10% fetal bovine serum. Inhibitors were added during the treatment period or during the recovery period. DTT was added only during the recovery period because simultaneous addition of DTT to LLC-PK1 cells will block IDAM covalent binding (37). Cytotoxicity was determined by the leakage of lactate dehydrogenase (37,38).
Northern Analysis, Electrophoretic Mobility Shift Assays, and Nuclear Run-on Assays-Poly(A) RNA was prepared by oligo(dT) affinity chromatography as described (38). Poly(A) RNA samples were size fractionated by denaturing agarose gel electrophoresis prior to transfer to nitrocellulose. cDNA probes were labeled with [ 32 P]dCTP (DuPont NEN) by random priming using a kit (Boehringer Mannheim). Human hsp70 cDNA (44) was obtained from American Type Culture Collection. The same blots were probed sequentially for hsp70 and ␤-actin, which served as an internal control.
For EMSA analysis, nuclear protein extraction was carried out as described (45). Cells were suspended in hypotonic lysis buffer followed by high salt extraction of the nuclei. Other procedures for EMSA analysis were performed according to Mosser et al. (46). Protein-DNA binding was carried out at 25°C for 20 min in a 12.5-l reaction mixture containing 5 g of nuclear proteins, 0.3 g of poly(dI-dC), and 80 pg of 32 P-labeled human HSE with a sequence identical to that used by Mosser et al. (46). In some cases, rabbit polyclonal antibody raised against murine HSF1, a gift of Dr. Rick Morimoto, was added (23). Shifted complexes were separated on a 4% polyacrylamide gels with 0.5 ϫ TBE as the running buffer.
Nuclear run-on experiments were performed using a standard procedure (47). Plasmids containing hsp70 and ␤-actin cDNAs were linearized, denatured, and immobilized on nylon membrane by slot blotting. Nuclei were incubated with 200 Ci of [ 32 P]UTP at 30°C for 30 min. Newly synthesized RNA was hybridized to the cDNA inserts on the nylon membranes, and the blots were exposed to DuPont Cronex film at Ϫ70°C with intensifying screens.
Immunoblotting And Immunofluorescence Analysis-Western blot analysis was carried out using monoclonal antibody (Amersham Corp.) specific for the stress-inducible 72-kDa member of the 70-kDa heat shock protein family (HSP70). Cells were lysed with SDS-gel sample preparation buffer, and proteins (20 g) were separated by SDS-polyacrylamide gel electrophoresis under reducing conditions. After proteins were transferred to nitrocellulose membrane, the membrane was blocked with 5% nonfat milk and probed with a 1:1000 dilution of anti-HSP70 antibody. After incubating with horse radish peroxidaseconjugated goat anti-mouse IgG, protein bands were detected by enhanced chemiluminescence using a kit (Amersham Corp.).
Immunofluorescence analysis was done essentially as described (23) using the anti-HSP70 monoclonal antibody or the anti-HSF1 rabbit polyclonal antibody used for EMSA analysis (see above). LLC-PK1 cells, grown to confluence on glass coverslips coated with bovine type I collagen, were treated with IDAM for 15 min, washed with PBS, and returned to normal medium. At the appropriate time, cells were washed with PBS, fixed in absolute methanol for 2 min at Ϫ20°C, and subsequently blocked with 1% bovine serum albumin in PBS for 20 min. Immunostaining was performed by incubating coverslips with either monoclonal anti-HSP70 diluted 1:500 or polyclonal anti-HSF1 antibody diluted 1:300. Indirect immunofluorescent detection was achieved using dichlorotriazinylaminofluorescein-conjugated goat anti-mouse or antirabbit IgG (Jackson ImmunoResearch, West Grove, PA), both diluted 1:300. All antibodies were diluted in PBS containing 1% bovine serum albumin. Coverslips were mounted and observed with a Nikon episcopic fluorescence microscope using a 40ϫ objective.
Protein and Nonprotein Thiol Assays, Covalent Binding, and Protein Cross-linking Analysis-Protein (PSH) and nonprotein thiols were determined using Ellman's reagent (48) with a minor modification for determination of the PSH (49). Cells were solubilized in 0.1% deoxycholate and deproteinized with cold trichloroacetic acid at a final concentration of 5%. Protein-free supernatants were collected for analysis of nonprotein thiols. For PSH assays, proteins were precipitated with 5% trichloroacetic acid, the pellet was washed twice with the same solution, and the denatured proteins were solubilized in 0.5% SDS by sonication before an aliquot was used to determine PSH.
Covalent binding of [ 14 C]IDAM to cellular macromolecules was determined as described (38) with modification. Confluent cultures in 12-well dishes were exposed to 75 M [ 14 C]IDAM (specific activity, 20 Ci/mol) for 15 min in 1 ml of EBSS. Covalent binding was determined immediately or after cells were washed with PBS and allowed to recover in medium for 15 min in the presence or the absence of 10 mM DTT. Cells were washed with PBS and fixed directly in the wells with 10% trichloroacetic acid. Fixed cells were washed in the dishes with 10% trichloroacetic acid before the proteins were solubilized in 0.5 ml of 0.1 N KOH. Radioactivity in neutralized samples was determined by liquid scintillation spectrometry.
Protein cross-linking was determined by the formation of high molecular weight complexes basically as described (40). Cells were labeled to steady state with [ 35 S]methionine and [ 35 S]cysteine (1 Ci/ml) for 24 h in methionine-and cysteine-free DMEM supplemented with 5% (v/v) DMEM (to allow for enough cysteine and methionine to support protein synthesis) and 10% fetal bovine serum. After treatment, cells were solubilized in SDS sample preparation buffer with or without ␤-mercaptoethanol and heated at 95°C for 15-20 min. Equal amounts of radiolabeled samples were loaded onto 6% SDS-polyacrylamide gels for electrophoresis under nonreducing conditions. Formation of high molecular weight complexes in dried nonreducing gels was visualized by autoradiography.
Quantitation of Autoradiograms and Immunoblots-The data from Northern blot, nuclear run-on, and Western immunoblot analyses were quantitated by densitometric scanning using a Bio Image Densitometer (Bio Image, Ann Arbor, MI). The amount of hsp70 mRNA on Northern blots was normalized by taking the ratio of the hsp70 value relative to the ␤-actin signals in each lane; data are expressed as the fold increase in hsp70 mRNA relative to untreated cells within each experiment. The nuclear run-on data were quantitated by normalizing the hsp70 signal to the signal for ␤-actin within each treatment. The normalized data are expressed as the fold increase in newly transcribed hsp70 mRNA in nuclei from IDAM-treated cells relative to nuclei from untreated cells. The fold increase in HSP70 protein expression on immunoblots was determined by comparing the data from treated cells with data from untreated cells.

RESULTS
Induction of hsp70 Transcription, HSF DNA Binding, hsp70 mRNA, and HSP70 by IDAM-We used the porcine renal epithelial cell line, LLC-PK1, for these studies because the cytotoxic characteristics of IDAM have been well characterized in these cells (37). The sulfhydryl alkylating agent IDAM induces HSPs, including HSP70, in chick embryo and human cells (42). Pulse treatment of LLC-PK1 cells with different concentrations of IDAM increased hsp70 mRNA in a concentration-dependent manner (Fig. 1B). After a 15-min treatment with 30 M IDAM, hsp70 mRNA increased within 1 h, peaked at 2 h, and declined at later time points (Fig. 1A). Induction of HSP70 protein expression was maximal at 6 h, lagging behind the maximal induction of hsp70 mRNA (Fig. 2). To determine the role of transcriptional events in IDAM-induced hsp70 mRNA accumulation, nuclear run-on analysis was performed. The rate of hsp70 transcription increased 16-fold compared with untreated cells (data not shown; see below). Therefore, IDAM treatment activated transcription of the hsp70 gene and synthesis of HSP70 protein in LLC-PK1 cells.
Regulation of hsp70 transcription is controlled by the binding of HSFs to HSEs located upstream of the hsp70 gene (21). To determine if HSF DNA binding was activated by IDAM, EMSA analysis was performed using the human HSE as a target. IDAM induced a concentration-dependent increase in nuclear HSE binding activity (data not shown). Mammalian cells contain a family of at least four HSFs (20). Supershift EMSA analysis demonstrated that HSF1 was the primary contributor to increased HSE binding activity in nuclei from IDAM-treated cells (Fig. 3).
HSF1 activation may involve translocation of HSF from the cytoplasm to the nucleus or relocalization within the nucleus (23). Immunofluorescence analysis detected some HSF1 in the cytoplasm, but the majority was in the nuclei of unperturbed cells (Fig. 4). This distribution was confirmed by immunoblotting soluble and nuclear fractions of LLC-PK1 cells (data not shown). There was no obvious change in the nuclear staining pattern of HSF1 in IDAM-treated and untreated cells. In contrast, diffuse HSP70 staining increased in the cytoplasm within 6 h after IDAM treatment (Fig. 4). Nuclear HSP70 was also visible, but the nucleolus did not stain markedly.
Effect of Inhibitors of Cytotoxicity and Protein Synthesis on hsp70 Expression-IDAM-induced cell killing of LLC-PK1 cells requires GSH depletion and lipid peroxidation (38). With prolonged IDAM treatment, the thiol reducing agent DTT, the iron chelator DFAM, which blocks hydroxyl radical formation, or the antioxidant DPPD prevent cytotoxicity and lipid peroxidation (38). After a 15-min treatment with IDAM, the addition of DTT, DFAM, or DPPD inhibited cytotoxicity (data not shown). In addition, the cell-permeable calcium chelator acetoxymethyl ester of EGTA also prevented IDAM-induced cell death indicating that increased cellular calcium may also play a role (data not shown), an effect not previously reported. However, only DTT prevented HSF binding (Fig. 5). DTT also abolished the IDAM-induced transcriptional activation of hsp70 (Fig. 6A) and HSP70 biosynthesis (Fig. 6B). Thus, IDAM-induced GSH depletion may activate HSF DNA binding activity, increasing hsp70 expression and HSP70 biosynthesis; covalent binding, increased cellular calcium, hydroxyl radical formation, and lipid peroxidation appear not to be involved.
It has been suggested that protein synthesis is involved in HSF activation because inhibitors of translation can prevent HSF DNA binding (29,34,46,50). Cycloheximide, at a concentration that inhibited over 90% of cellular protein synthesis, abolished HSF DNA binding in response to 75 M IDAM (Fig.  5), reduced hsp70 transcription (Fig. 6A), and decreased hsp70 mRNA accumulation (Fig. 7). Thus, ongoing protein synthesis is necessary for IDAM-induced hsp70 transcription.
DTT also inhibits protein synthesis (51), raising the possibility that the effect of DTT on HSF1 activation is due to a translational block. Indeed, a 1-h treatment with 10 mM DTT inhibited protein synthesis markedly (Table I). However, the ability of DTT to block HSF1 activation could be differentiated from the inhibition of protein synthesis in two ways. First, treatment of cells with 1 or 10 mM DTT for 15 min or less blocked HSF1 activation but did not inhibit protein synthesis ( Fig. 8; cf. Table I). Second, DTT still blocked HSF activation after treatment with high concentrations of IDAM, but cycloheximide was effective only at lower IDAM concentrations (Fig. 9).
Both DTT and protein synthesis inhibitors have been reported to block HSF binding following heat shock (29,34,50,52). We confirmed that cycloheximide prevented HSF activation by modest (43°C) hyperthermia in LLC-PK1 cells (data not shown). We also compared the effects of DTT and cycloheximide on HSF activation by thermal stress. In one experiment, DTT blocked the induction of HSF DNA binding following 43°C heat shock but only at concentrations that inhibited protein synthesis; DTT did not prevent activation of HSF by severe (46°C) heat shock (Fig. 10 cf. Table I). In additional experiments, DTT had no effect on heat-induced activation of HSF1 DNA binding (data not shown). Taken together, the data suggest that HSF activation by IDAM and heat shock occur by distinct mechanisms. In LLC-PK1 cells, HSF1 activation by heat shock takes place in the apparent absence of a redox perturbation that is inhibited by DTT.
The Role of Protein Thiol Oxidation and Protein Aggregation in HSF1 Activation-Proteotoxic damage due to covalent binding of electrophiles to protein is believed to be an important activator of HSF (6). However, depletion of GSH can also lead to modification of protein structure (39); therefore, we measured cellular thiols and alkylation of proteins after IDAM treatment.  romolecules (Table II), whereas nonprotein thiols, i.e. GSH, decreased from 12 nmol/mg protein to 2 nmol/mg, a decrease of over 80% (Table II). Thus, the decrease in PSH (22 nmol/mg) cannot be accounted for by acetamidylation (184 pmol/mg) of cysteinyl residues. However, when IDAM-treated cells were washed with 10 mM DTT under conditions that block HSF activation but not protein synthesis, the loss of PSH was reversed without decreasing IDAM binding (Table II). Thus, covalent binding was dissociated from HSF activation.
PSH oxidation can occur through formation of PSSP or by formation of mixed disulfides with small molecules such as oxidized glutathione to form a protein-glutathione mixed disulfide (PSSG). Intermolecular PSSP would appear as large complexes after nonreducing SDS-polyacrylamide gel electrophoresis, whereas PSSG mixed disulfides would not (40). When LLC-PK1 cells were labeled to steady state with [ 35 S]methionine and [ 35 S]cysteine and then treated with IDAM for 15 min, we observed high molecular weight complexes that failed to migrate into SDS-gels under nonreducing conditions, thus IDAM treatment caused concentration-dependent increase in PSSP formation from mature proteins (data not shown). Approximately 6 -10% of the total label did not migrate into the gel (data not shown; n ϭ 2). More importantly, washing IDAMtreated cells with DTT for only 15 min prevented PSSP formation up to 60 min after IDAM treatment (Fig. 11). Aggregates appearing at the top of nonreducing gels were also dissociated by DTT after proteins had been extracted from the cells (data not shown). Therefore, IDAM-induced loss of to GSH and PSSP formation correlates with activation of HSF1 DNA binding activity. DISCUSSION Current models suggest that hyperthermic protein denaturation activates the heat shock response (6, 29) by disrupting a feedback loop through which HSPs directly or indirectly suppress HSF1 activation (29 -31). It has been suggested that alkylating agents activate the heat shock response because they damage proteins by modifying functional groups covalently (33,34). However, alkylating agents, including IDAM, also cause secondary cytotoxic signals, such as depletion of glutathione, increased cellular calcium, oxidative stress and lipid peroxidation (37,53), which could also activate HSF DNA binding (39, 54 -56). We have been probing the role of these cytotoxic signals in activation of stress-responsive genes, including hsp70 (36 -38, 57).
IDAM activated HSF1 and induced a classic heat shock response in LLC-PK1 cells. However, increased cellular calcium, reactive oxygen species, and lipid peroxidation were not involved. On the other hand, DTT addition blocked HSF1 activation without altering IDAM covalent binding. DTT also reversed the loss of PSH and PSSP formation. Thus, the impor-

H]leucine incorporation into trichloroacetic acid insoluble cellular material. [ 3 H]Leucine
(3 Ci/well) was added to confluent cultures of LLC-PK1 cells in 12-well dishes in 1 ml of DMEM (final specific activity, 3.6 Ci/mol) in the absence or the presence of DTT. After 2 h, the samples were collected, and protein synthesis was assayed essentially as described (38). tant signal for IDAM-induced HSF1 activation is GSH depletion and subsequent PSSP formation, not covalent binding. Conceivably, loss of PSH could result from formation of PSSG mixed disulfides after oxidation of cellular GSH to oxidized GSH. However, on a molar basis the loss of GSH (10 nmol/mg protein), could not account for the loss of PSH (22 nmol/mg protein) even if all the GSH lost appeared as oxidized GSH (i.e. 5 nmol/mg). Because IDAM binding to thiols is facile, it is more likely that the GSH is lost due to thioether formation; PSSP formation occurs as a result of a secondary oxidative stress. Regardless, the data suggest that PSSP formation contributes significantly to the loss of PSH and activation of HSF1. The mechanisms through which PSSP formation activates HSF1 is not clear, but a plausible model can be constructed based on current models (Fig. 12). The model depicts HSP70 acting as a negative regulator of HSF1, as suggested for the heat shock cycle (21). At the same time, HSP70 recruited from a "free pool" assists in folding nascent polypeptide chains (4), thus establishing an equilibrium between HSF1 and HSP70. Following IDAM treatment, disulfide cross-linked and malfolded proteins formed as a result of GSH depletion sequester HSP70 releasing HSF1. Although this simple model emphasizes the role of HSP70, it is clear that activation of HSF1 is a complex process involving multiple intermolecular and intramolecular points of regulation including phosphorylation (58,59). However, the model is supported by the data and is consistent with studies showing that HSF is activated by diamide, an agent  Table I).
In multiple repeat experiments, treatment with DTT (1-10 mM) for various times had no effect on HSF1 activation by heat shock. C]IDAM and the levels of cellular non-PSH and PSH were determined as described under "Experimental Procedures." The data are the mean Ϯ S.D. from three independent experiments (n ϭ 3). Significant difference (p Ͻ 0.05) among PSH means were determined by ANOVA followed by multiple comparison using least significant differences. Means with a common letter designation (a or b) are not different from other means with the same designation. Significant differences (p Ͻ 0.05, asterisk) between the means within the non-PSH or IDAM binding data sets were determined using Student's t-test. that oxidizes GSH and denatures cellular protein (39).
For several reasons, we focused our attention on changes in the protein thiol-disulfide redox status rather than reactive oxygen species as a trigger for HSF1 activation. First, PSSP formation and the loss of PSH were reversed by DTT treatments that diminish HSF1 activation. Second, loss of PSH occurs concomitantly with loss of GSH and formation of PSSP. Third, chelating cellular iron to prevent hydroxyl radical formation via the Fenton reaction does not block HSF1 activation, thus excluding a role for oxidative damage to proteins by hydroxyl radical (60). However, the source of the oxidants that lead to PSSP formation is not apparent. Although autooxidation is possible, increased superoxide and hydrogen peroxide formation due to IDAM treatment could also play a role in protein disulfide formation. For example, in the absence of adequate GSH to support glutathione peroxidase activity, formation of superoxide as a side product of cellular metabolism and subsequent dismutation to hydrogen peroxide via superoxide dismutase could yield an increasing cellular burden of hydrogen peroxide. Hydrogen peroxide could directly oxidize thiols (Equation 1) as an intermediate step to PSSP formation (Equations 2 and 3; Refs. 61 and 62).
Likewise, PSH could also react with superoxide to form protein disulfides (Equation 4). Finally, oxidized GSH accumulation could lead to formation of mixed disulfides of protein and glutathione (PSSG; Equation 5).
Attack of a second PSH would yield a net oxidation of protein thiols to a mixed protein disulfide (Equations 6 and 7).
Any of these mechanisms could play a role in activation of HSF1 by IDAM.
Although the model suggests that PSH oxidation indirectly activates HSF1, a second possibility that must be considered is that oxidative stress may alter DNA binding activity by modifying HSF1 directly. Oxidation of cysteinyl thiols can alter DNA binding activity negatively, in the case for FOS-JUN heterodimers (63), or positively, as with the prokaryotic OxyR transcription factor (64). However, the 2 mM DTT present in the binding buffer did not block DNA binding activity of HSF nor did DTT treatment of the cells block heat-induced activation of HSF1. Therefore, HSF1 can be activated by heat and IDAM in the absence of DTT-sensitive modification of HSF1. However, DTT does block HSF1 activation by heat in HeLa cells (52), thus different cells may have redox-dependent and -independent mechanisms of HSF activation after heat shock.
The model in Fig. 12 also explains the effect of cycloheximide on HSF1 activation by moderate concentrations of IDAM. Cycloheximide blocked HSF1 activation and hsp70 transcription by low concentrations of IDAM, suggesting that ongoing protein synthesis plays a role in agreement with other studies (29,34,46,50). It is likely that blocking protein synthesis with cycloheximide releases HSPs from folding nascent polypeptides, thus increasing the HSP pool available to handle the malfolded proteins obviating the need for HSF1 activation (15,29). Alternatively, the newly synthesized polypeptide pool may be more susceptible to alkylating damage (34,50). However, we favor the former hypothesis for two reasons. First, mature proteins, i.e. labeled to steady state with [ 35 S]methionine/cysteine, formed disulfide-linked aggregates even following treatment with low concentrations of IDAM. The fact that as much as 6 -10% of the steady state labeled proteins appear in the PSSP aggregates argues that mature proteins are targets for protein thiol oxidation. Second, regardless of whether newly synthesized proteins are more susceptible to covalent attack, binding does not appear to be a signal for HSF1 activation. However, elucidating the detailed mechanism underlying the role of nascent peptides and protein synthesis in HSF1 activation by alkylating agents requires further study.
Because DTT inhibits protein synthesis (51), it could be argued that like cycloheximide, all or part of the DTT effect on HSF activation was due to inhibition of protein synthesis. This appears not to be the case for several reasons. First, DTT prevented HSF DNA binding and hsp70 expression under conditions that did not affect protein synthesis. Second, DTT was effective at concentrations of IDAM at which cycloheximide had no effect. Third, concentrations of DTT that had no effect on protein synthesis did not diminish HSF1 activation by moderate heat shock, whereas cycloheximide blocked HSF1 activation. Therefore, the effects of DTT and cycloheximide on HSF1 activation by IDAM can be segregated in LLC-PK1 cells.
Interestingly, HSF1 had a predominantly nuclear staining pattern in LLC-PK1 cells; the pattern was not altered markedly following IDAM treatment. This is in contrast to the heat shock response in other mammalian cells in which HSF1 translocates to the nucleus from the cytoplasm or relocates within the nucleus (23). However, a nuclear localization similar to that observed in LLC-PK1 cells is also observed in unperturbed Drosophila SL2 cells (26). Likewise, no characteristic nucleolar translocation of HSP70 was detected upon IDAM treatment, again in contrast to cellular HSP70 distribution after heat shock in other cells (23,65). Thus, there may be characteristic differences in HSF1 and HSP70 subcellular distribution between LLC-PK1 and other mammalian cell lines and/or between heat shock and alkylation stress.
In summary, although covalent modification of cysteinyl thiols by IDAM alters protein structure, it appears that hsp70 transcription is due to the secondary depletion of cellular GSH. The loss of GSH allows oxidation of protein thiols to PSSP. The formation of PSSP appears to be linked to HSF1 activation and increased transcription of hsp70. It will be important to determine if this is a general mechanism of HSF1 activation by reactive chemical toxicants.