Intramolecular Repression of Mouse Heat Shock Factor 1

Abstract

The pathway leading to transcriptional activation of heat shock genes involves a step of heat shock factor 1 (HSF1) trimerization required for high-affinity binding of this activator protein to heat shock elements (HSEs) in the promoters. Previous studies have shown that in vivo the trimerization is negatively regulated at physiological temperatures by a mechanism that requires multiple hydrophobic heptad repeats (HRs) which may form a coiled coil in the monomer. To investigate the minimal requirements for negative regulation, in this work we have examined mouse HSF1 translated in rabbit reticulocyte lysate or extracted from Escherichia coli after limited expression. We show that under these conditions HSF1 behaves as a monomer which can be induced by increases in temperature to form active HSE-binding trimers and that mutations of either HR region cause activation in both systems. Furthermore, temperature elevations and acidic buffers activate purified HSF1, and mild proteolysis excises fragments which form HSE-binding oligomers. These results suggest that oligomerization can be repressed in the monomer, as previously proposed, and that repression can be relieved in the apparent absence of regulatory proteins. An intramolecular mechanism may be central for the regulation of this transcription factor in mammalian cells, although not necessarily sufficient.

The increased synthesis of heat shock proteins (HSPs) is a response of cells of many, if not all, organisms to temperatures above normal and to diverse physiological and experimental stress stimuli (14, 25). In eukaryotes, the induction of HSP-encoding genes is regulated at the transcriptional level by heat shock factor (HSF), which binds multiple copies of an upstream sequence, the heat shock element (HSE), consisting of contiguous 5-bp modules (nGAAn) in alternating orientations (12, 26).

HSFs from a broad range of species are characterized by a conserved DNA-binding motif in the amino terminus and adjacent hydrophobic heptad repeats (HR-A and HR-B [HR-A,B]) which mediate subunit trimerization via an α-helical coiled-coil structure (17, 34, 41, 43, 46). A carboxy-terminal hydrophobic heptad repeat (HR-C) is also found in many members of this transcription factor family (26, 36, 49).

In vertebrates, which contain multiple HSFs encoded by distinct genes (30, 31, 49), the transcriptional response to heat stress is mediated by HSF1. This protein is constitutively synthesized and mostly held in the nucleus in the apparent form of a monomer modified by phosphorylation (10, 21, 30, 31, 49). The heat shock stimulus rapidly activates the DNA-binding function of HSF1 by a reversible step of subunit trimerization (3, 5, 20, 33, 36, 43, 47–50), while a distinct step enables the function of a constitutively active transcriptional activator domain in the carboxy terminus (16, 19, 21, 32, 42, 54).

Previous studies have shown that the carboxy-terminal hydrophobic repeat (HR-C) is required to repress trimerization of HSF1 in human cells at physiological temperatures, and a similar requirement was found for the HSF of Drosophila melanogaster (36). Furthermore, deletions or substitutions of hydrophobic residues in either HR-C or trimerization (HR-A,B) domains caused constitutive oligomerization and DNA-binding activity of human HSF1 in Xenopus oocytes in which the exogenous HSF adopted the host cell induction temperature (53). This led to the proposal that trimerization is repressed in the monomer by coiled-coil interactions which may be stabilized by other domains of the protein (33, 36) and also by other factors, possibly the 70-kDa heat shock protein HSP70 (2, 9, 29, 53). A spontaneous activation often observed during overexpression of HSF1 in transfected mammalian cells (13, 36, 40) and the constitutive oligomerization and activity of HSFs expressed as recombinant proteins in Escherichia coli (7, 11, 22, 26, 27, 35, 49) also suggested the action of a limiting inhibitory molecule. Moreover, in vitro experiments showed activations of HSFs by temperature and conditions that affect protein conformation, including acidic pH, in cell extracts or reticulocyte lysates (2, 23, 28, 39, 47, 52).

To examine the minimal requirements for repression, in this study, mouse HSF1 was translated in rabbit reticulocyte or extracted and purified after limited expression in E. coli. We show that in these systems an apparent monomer can be converted to the active form by increases in temperature or mildly acidic pH even after purification, while mutations of either heptad repeat region allow constitutive oligomerization and HSE binding. Furthermore, purified mouse HSF1 was activated proteolytically, most probably by excision of fragments able to oligomerize. While chaperone activities of reticulocyte lysate and E. coli may assist the monomeric folding of HSF1, our results suggest a mechanism of repression in the monomer (36, 53).

MATERIALS AND METHODS

Constructs for HSF expression.

Constructs based on plasmid pET3b (44) were used for expression of murine HSF1 (503 amino acids [39]) and of mutant HSFs in rabbit reticulocyte lysate, E. coli, and E. coli S30 extracts. DNA sequencing was done by the dideoxynucleotide chain termination method; other procedures were as described in a cloning manual (38). PCR was done with Vent DNA polymerase (New England Biolabs). The codon substitution in HSF1[H179R] was made by PCR with oppositely oriented primers introducing G at position 536 and C at position 537, which destroys a SphI site, engineers a BssHII site, and replaces the codon for histidine 179 with an arginine codon. The primer pairs used were as follows: primers 1 and 2, 5′-CCTCGTGAGCGACCCGGACACA-3′ (position 210; sense) and 5′-TGCTGCTGGGCGCGCTTCTG-3′ (position 687; antisense), respectively; and primers, 3 and 4, 5′-AGAAGCGCGCCCAGCAGCAA-3′ (position 669; sense) and 5′-GGGGCTTGGGCTCCGGTTGTG-3′ (position 1212; antisense), respectively. After restriction and gel purification, fragments obtained with primers 1 and 2 and primers 3 and 4 and the plasmid were ligated. Constructs encoding HSF1[1-430] (containing amino acids 1 to 430 of HSF1) and HSF1[1-382] were generated by PCR using oppositely oriented primers corresponding to codons 443 to 449 (5′-ACGGCTCAGCCTCTGTTCCT-3′) and a primer replacing codon 431 or 383 with nonsense codons (5′-GGTTTCAGTTCTCTGCCTCAATAGG-3′ and 5′-CATCCAGCTAATCACTTAGCTCGTT-3′, respectively). The above constructs were sequenced. The other mutants with carboxy-terminal deletions analyzed in reticulocyte lysate were synthesized by in vitro transcription-translation of PCR products; the primers used were T7 promoter primer and each antisense primer with a mismatch engineering a stop codon (positions 1 to 503, 5′-AGAGCTCTAGGAGACAGTG-3′; positions 1 to 466, 5′-AGCTAAAAGAGCACAGGCAGCT-3′; positions 1 to 407, 5′-GCTCAGGTGTCCACACTGAAGCC-3′; positions 1 to 399, 5′-GTGTCATGTCAGCATGGTCTGCA-3′; positions 1 to 393, 5′-TGCAGTTAGTCCAGGTTGGAG-3′; positions 1 to 346, 5′-TAGGGGCTTAGTTTGAGGCAGCA-3′; positions 1 to 297, 5′-CTCTTGCTAGACACGGACCAGA-3′; positions 1 to 266, 5′-GATTTAGGAGATTATGGGTCCAGAG-3′; positions 1 to 228, 5′-CAGGGACTACTGTCGACCATACTTG-3′; positions 1 to 194, 5′-TGCACCAGTCAGATCAGGAACTGA-3′; and positions 1 to 191, 5′-TCAGTTACTGAATGAGCTTGTTG-3′).

In vitro translations and analysis of HSFs.

Translation reaction mixtures contained either reticulocyte lysate (70%) or S30 extract (30%), RNA (20 μg ml⁻¹) transcribed with T7 RNA polymerase, amino acids (20 μM each), [³⁵S]methionine (∼0.4 μCi ml⁻¹; 1,200 Ci mmol⁻¹; NEN), and 500 U of RNasin (Promega) ml⁻¹. Reactions (20-μl reaction volume) were done at 30°C for 60 min before addition of cycloheximide (10 μg ml⁻¹) to reticulocyte lysates or chloramphenicol (10 μg ml⁻¹) to S30 extracts. To calculate the yield of in vitro-translated HSF1, bands of interest were excised from dried sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels and dissolved in 30% H₂O₂ at 65°C for 3 h, and radioactivity was measured in scintillation fluid; background radioactivity was counted in a similar area above the band of interest.

For protein cross-linking, the reticulocyte lysate was centrifuged at 100,000 × g, and the supernatant (5 μl) was adjusted to 100 μl with buffer (100 mM triethanolamine [pH 8.0], 2 mM EDTA, 5% glycerol) and incubated for 30 min at 25°C in the presence of ethylene-glycol-bis-(succinimidylsuccinate) (EGS) (Pierce) at the concentrations indicated below; after the reaction was quenched with 150 mM glycine for 5 min, samples were adjusted to 500 μl with 10 mM Tris-Cl (pH 8.0)–140 mM NaCl–0.1% Nonidet P-40 and incubated with anti-HSF1 rabbit antiserum (1:5,000) (13). Immunocomplexes collected with Staphylococcus aureus (Sigma) were subjected to SDS-PAGE in 5 to 15% polyacrylamide gradient gels. For cross-linking of HSF1[H179R] purified from E. coli extracts, ∼30 ng of protein was incubated with EGS as described above and directly electrophoresed after the reaction was quenched. For gel filtrations with the SMART system (Pharmacia), supernatants (100,000 × g) (40 μl) were injected into a Superose 6 PC 3.2/30 column calibrated with blue dextran, acetone (initial volume = 0.8 ml; total volume = 2.14 ml), and the following proteins: carbonic anhydrase (27 kDa), ovalbumin (43 kDa), bovine serum albumin (BSA) (66 kDa), alcohol dehydrogenase (150 kDa), beta-amylase (200 kDa), apoferritin (443 kDa), and thyroglobulin monomer (669 kDa) and dimer (1,338 kDa). Chromatography was done at ambient temperature (∼25°C) in 20 mM Tris-Cl (pH 7.4)–140 mM NaCl–1 mM EDTA–10% glycerol–0.2% Triton X-100–0.5 mM dithiothreitol (DTT)–protease inhibitors as described below. Fractions were analyzed by SDS-PAGE followed by autoradiography or semidry blotting onto nitrocellulose and enhanced chemiluminescence (ECL) detection with anti-HSF1 (1:5,000) and peroxidase-coupled donkey anti-rabbit antibody (1:10,000). Densitometric analysis of ECL films was done with a Molecular Dynamics scanner.

Heat shock was done in 0.6-ml thin-wall tubes in a thermal cycler (MJ Research, Inc.). DNA-binding assays were done in a 20-μl volume containing 3 to 5 μl of reticulocyte lysate or E. coli extract (the concentration ranges of HSFs were 0.4 to 0.6 μg ml⁻¹ in the lysates and 1.0 to 2.0 μg ml⁻¹ in the extracts) or chromatographic fraction. Binding mixes contained 0.4 mg of poly(dI-dC)-poly(dI-dC) ml⁻¹, 20 mg of BSA (fraction V; Sigma) ml⁻¹, and 200 to 500 fmol of ³²P-HSE (upper strand, 5′-GTCGACGGATCCGAGCGCGCCTCGAATGTTCTAGAAAAGG-3′; underlining indicates the HSE sequence) labeled with Klenow enzyme as described previously (51). Binding reactions were done in melting ice for 10 min, and then electrophoresis in 1% agarose gel at ambient temperature was done as described elsewhere (51). Gels blotted onto DE81 paper were autoradiographed and analyzed in a PhosphorImager; the radioactivity of excised bands was measured in a scintillation counter.

Expression of HSFs in E. coli, purification, and mild proteolysis.

E. coli BL21(DE3) pLysS (44) carrying a pET3b derivative coding for HSF1, HSF1[1-382], HSF1[1-430], or HSF1[H179R] was grown in Luria-Bertani medium, containing both ampicillin (125 μg ml⁻¹) and chloramphenicol (25 μg ml⁻¹), by vigorous shaking at 37°C until an optical density at 600 nm of 0.7 was reached, at which point incubation was continued in the presence of 0.4 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 10 min unless otherwise indicated. After being pelleted at 8,000 × g for 20 min, the cells were resuspended in buffer (20 mM HEPES [pH 7.5], 0.5 mM DTT) and subjected to two freeze-thaw cycles for lysis by the T7 phage lysozyme encoded by the pLysS plasmid (44). The lysates were centrifuged at 100,000 × g for 30 min at 4°C, and the supernatants were quick-frozen and stored at −86°C. Cells carrying the expression plasmid were counted after serially diluted culture aliquots, withdrawn prior to IPTG treatment, were plated on ampicillin-containing medium. To calculate the number of molecules per cell, the HSF1 content of S100 extracts was compared in immunoblots with overproduced HSF1 purified by HSE affinity chromatography (13), quantitated by Coomassie blue staining of SDS gels by use of a BSA standard. After 10 min of IPTG treatment, 20 to 40 ng of each HSF was extracted from 10⁹ ampicillin-resistant cells, corresponding to 2 × 10² to 4 × 10² molecules per cell (theoretical HSF1 molecular mass = 55 kDa). Extraction in the presence of SDS did not significantly increase the recovery of HSFs. For each purification round, E. coli extracts equilibrated in buffer (20 mM sodium phosphate [pH 7.1], 2 mM EDTA, 10% glycerol, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, 1 μg each of leupeptin and pepstatin ml⁻¹, 5 μg of aprotinin ml⁻¹) were chromatographed on Mono S HR5/5, and proteins were eluted with a linear NaCl gradient (0.05 to 1.0 M). Eluates from three Mono S runs (corresponding to 50 mg of starting material) were pooled and sequentially fractionated on Blue Sepharose CL-6B HR5/5 that was step eluted with 0.6 M NaCl in buffer (10 mM Tris-Cl [pH 7.4], 2 mM EDTA, 10% glycerol, 0.5 mM DTT) and on Mono Q PC 1.6/5 that was eluted with a linear NaCl gradient (0.15 to 0.4 M). At each step, fractions of interest were identified by immunoblotting with rabbit anti-HSF1 serum (13) preadsorbed on E. coli BL21 (DE3) proteins or with immunoglobulins affinity purified on protein A-Sepharose and HSF1 coupled to CNBr-activated Sepharose 4B (Pharmacia) (10). Each purification yielded approximately 1.0 μg of HSF1 that was nearly homogeneous (80%) and at a concentration of 2.0 μg ml⁻¹ in Mono Q fractions 5 and 6. Similar procedures were employed for purifications of HSF1[H179R] and HSF1[1-430], except that proteins were applied to Blue Sepharose in 0.3 M NaCl and eluted with linear NaCl gradients and the Mono Q step was omitted for HSF1[1-430] purification. Incubations of purified proteins at the indicated temperatures or pH or with proteases were typically done with 6 μl (approximately 12 ng) diluted threefold in buffer (70 mM potassium glutamate, 20 mM sodium phosphate at the indicated pH values, and 3.0 mM DTT). Samples were layered on mineral oil or on Parafilm to reduce protein loss. The pH values were measured with a pH meter probe in parallel samples containing buffers mixed in the appropriate proportions. Mild proteolysis with chymotrypsin or S. aureus protease V8 was done at pH 7.2 for 5 min at 20°C, followed by addition of aprotinin or Nα-p-tosyl-l-lysine chloromethyl ketone (TLCK) (Sigma); the protease/HSF1 ratios (wt/wt) were as described below (see the legend to Fig. 5).

FIG. 5.

Proteolytic activation of purified HSF1. (A) Electrophoretic mobility shift assays of HSF1 (Mono Q fraction 5) after incubation with chymotrypsin at the indicated concentrations (upper panel). Incubations were done at 20°C for 5 min followed by addition of aprotinin. The ratios (wt/wt) of protease to HSF1 were 1:50, 1:14, 1:5, 1:1.66, and 1:0.55 (lanes 4 to 8). In the samples in lanes 2 and 3, the chymotrypsin concentration is the same as in lane 8 except that aprotinin was added together with the protease. The smear above the free ³²P-HSE is due to aprotinin. Lanes 9 and 10 show parallel incubations at 20 or 43°C for 5 min without aprotinin. The shifted counts in lane 7 were 1/10 of the counts in lane 10, as determined by PhosphorImager analysis. Aliquots of samples 1 to 8 were analyzed by SDS-PAGE followed by immunoblotting (lower panel). B and F, bound and free ³²P-HSE, respectively. (B) Superose 6 gel filtration of chymotryptic digests. After addition of aprotinin, the digests (similar to samples 4 to 8 in panel B) were pooled and subjected to gel filtration, and the fractions were analyzed by immunoblotting. The positions of major proteolytic fragments (a to c) are indicated.

RESULTS

Inducible oligomerization of mouse HSF1 in reticulocyte lysate.

The oligomerization step associated with activation of DNA binding of in vitro-translated mouse HSF1 (13, 39) was monitored by protein cross-linking with EGS (1). Figure 1A shows that the 70-kDa monomer was predominant in the lysate that was kept at the translation temperature (30°C), while a complex with a molecular mass of >200 kDa was formed in the lysate incubated at 43°C prior to cross-linking. We found that after induction at 43°C, the DNA-binding activity was stable for hours at 30 or 25°C, although the amount of cross-linkable protein was not determined (10). It should be noted that reticulocyte lysates do not contain a detectable endogenous HSE-binding activity (39).

FIG. 1.

Temperature-dependent oligomerization of mouse HSF1 and properties of mutant proteins translated in reticulocyte lysate. The concentration range of HSF1 is 0.4 to 0.6 μg ml⁻¹. (A) [³⁵S]methionine-labeled mouse HSF1 that was kept at the translation temperature or incubated at 43°C for 20 min was cross-linked at 25°C with EGS at the concentrations indicated above the lanes. Immunoprecipitation with anti-HSF1 was carried out after quenching as described in Materials and Methods. The cross-linking products (dots) and the monomer and molecular mass markers (arrows) are indicated. Counts at the tops of the lanes with 1.0 mM EGS were seen occasionally and may represent an artifact. (B) [³⁵S]methionine-labeled HSFs translated from templates produced by PCR as described in Materials and Methods. 503* was translated from plasmid C12A (39). Amino acid endpoints (Δ), predicted molecular weights (in thousands), and molecular mass markers (lane M) (arrows) are indicated. HSF1 migrates as two closely spaced bands due to partial phosphorylation, as indicated by phosphatase treatment (10). (C) Superose 6 gel filtrations of HSF1, HSF1[1-382], and HSF1[1-430] kept at the translation temperature (30°C) or incubated at 43°C for 20 min. HSF1[1-382] was recovered in the void volume after incubation at 43°C, most probably due to formation of large (inactive) aggregates (10). Peak elutions of thyroglobulin (669 kDa), beta-amylase (200 kDa), and BSA (69 kDa) are indicated above the lanes. Peak fractions 21 and 16 correspond to ∼100 and ∼520 kDa, respectively, consistent with previous reports showing that human and Drosophila HSFs behave as nonglobular proteins in gel filtration (24, 48). L. O., load on.

To find out whether the activity of mouse HSF1 is negatively regulated at the translation temperature, the coding sequence was deleted from the 3′ end and mutant proteins (Fig. 1B) were examined by gel filtration chromatography in a Superose 6 column and by electrophoretic mobility shift assays. Gel filtration experiments (Fig. 1C) showed that full-length HSF1, corresponding to the monomer in Fig. 1A, peaked mostly in fractions 20 to 22, while 60% eluted with a new peak in fractions 15 to 17 after incubation at 43°C. HSFs ending with F-466 (10) and with N-430 (Fig. 1C) behaved similarly in gel filtrations and also showed maximal DNA-binding activity at 43°C (Fig. 2A). In contrast, mutants lacking the HR-C region, ending with D-382 or N-346, displayed activity at the translation temperature (Fig. 2A) and behaved as oligomers, as shown for HSF1[1-382], eluting in gel filtration fractions 18 and 19 (Fig. 1C). Mutants with partial deletions of the HR-C region, ending with T-407 or T-399, displayed higher activity at 37 than at 43°C, and those ending with N-346, V-297, S-266, Q-228, or I-194 exhibited activity at the translation temperature and at 37°C but were inactivated at 43°C (Fig. 2A). The removal of three more amino acids from the trimerization domain in HSF1[1-191] caused substantial loss of activity at the translation temperature, and HSF1[1-156], lacking most of the trimerization domain, did not display activity in our assays (Fig. 2A). In gel filtrations, HSF1[1-194] (predicted molecular mass, 22 kDa) peaked as an apparent trimer in fractions 22 to 23, while HSF1[1-191] behaved as a monomer coeluting with carbonic anhydrase (27 kDa) in fractions 26 to 27 (Fig. 2B).

FIG. 2.

Properties of carboxy-terminally deleted HSFs and of HSF1[H179R] in reticulocyte lysate. Proteins were labeled with [³⁵S]methionine. (A) DNA-binding activities of HSFs incubated at the indicated temperatures for 20 min. For each HSF, the DNA-binding domain (DBD) (amino acids 16 to 120), trimerization domain (HR-A,B) (amino acids 137 to 212), and carboxy-terminal hydrophobic repeat HR-C (amino acids 378 to 407) are shown (boxes); boundaries of each domain are as in reference 39. Note that the loss of activity of many mutants incubated at 43°C is due to inactivations, but polypeptides were intact as determined by SDS-PAGE analysis (10). When the activity is compared with those of [³⁵S]methionine-labeled HSFs shown in Fig. 1B, it should be considered that HSF1[1-399] contains 11 methionines, like full-length HSF1, and that HSF1[1-382] and HSF1[1-346] contain 9 and 8 methionines, respectively. Note that the translation efficiency was higher for the shorter polypeptides. (B) Gel filtrations of HSF1[1-194] and HSF1[1-191]. Samples from separate gel filtrations were mixed prior to gel loading. The peak elution of carbonic anhydrase (27 kDa) is indicated. The carboxy-terminal amino acids and apolar residues of the trimerization domain B region (diamonds and dots, respectively) are indicated. (C) Properties of HSF1[H179R]. (i) Results of electrophoretic mobility shift assay after incubation at the indicated temperatures for 20 min. (ii) HSF1[H179R] and HSF1 analyzed by SDS-PAGE. (iii) HSF1[H179R] cross-linked with EGS. Upon incubation at 43°C, HSF1[H179R] was poorly recovered after the high-speed centrifugation routinely performed prior to cross-linking or gel filtration; when the centrifugation step was omitted, the protein was immunoprecipitated inefficiently, suggesting that it becomes part of aggregates. B and F, bound and free ³²P-HSE, respectively.

These results show that in the lysate, as in vivo, the carboxy-terminal portion is required to inhibit oligomerization and DNA-binding activity and that the HR-C region is critical for the inhibition (30, 36, 53).

In a previous study, Zuo et al. (53) showed that small deletions or amino acid substitutions in the trimerization domain (HR-A,B) deregulate human HSF1 expressed in Xenopus oocytes. We examined in reticulocyte lysate a mutant bearing a substitution of arginine for histidine 179 in the trimerization domain. A histidine in this position is conserved in HSFs of budding yeasts, Drosophila, and vertebrates and is in phase with the hydrophobic heptad repeats (34, 49). In SDS gels, the in vitro-translated HSF1[H179R] comigrated with HSF1 (Fig. 2C, panel ii). HSF1[H179R] displayed activity at the translation temperature (Fig. 2C, panel i) and was cross-linked as dimers and trimers at the translation temperature (Fig. 2C, panel iii; in gel filtrations, HSF1[H179R] repeatedly eluted in a broad range of fractions [10]). Following incubations of the lysate at 43°C, HSF1[H179R] was mostly inactivated and was poorly recovered (reference 10; see the legend to Fig. 2C). These results show that histidine 179 is required to inhibit the activity at the translation temperature, suggesting an inhibitory function of the trimerization domain also in the lysate.

Regulation of DNA binding of mouse HSF1 extracted from E. coli.

To find out whether the in vitro regulation observed in the experiments described above depends critically on constituents of reticulocyte lysate, HSF1 and three selected mutants were also examined in a prokaryotic system. For this, we used an IPTG-inducible expression system in E. coli (44) and extracted the HSFs after expression in each case of 200 to 400 molecules per cell. For this, E. coli cells were harvested after 10 min of IPTG treatment (see Materials and Methods). Tests done with whole-cell extracts showed that the DNA-binding activity of HSF1 or HSF1[1-430] was not detected at the ambient temperature (25°C) but was induced by incubation at 37 or 43°C for 20 min, whereas HSF1[1-382] and HSF1[H179R] displayed maximal activity at the ambient temperature (Fig. 3A, upper panels). Like in reticulocyte, HSF1[H179R] was inactivated at 43°C in these extracts (Fig. 3A, lane 12). Furthermore, similar results were obtained when these HSFs were translated in a commercial E. coli S30 extract (10). The results of gel filtrations shown in Fig. 3B demonstrate that the elution of HSF1 was shifted from fractions 20 to 22 to fractions 15 and 16 following incubation at 43°C. In contrast, HSF1[1-382] and HSF1[H179R] both behaved as oligomers at the ambient temperature, peaking around fractions 16 and 19, respectively. It should be noted that inactive and active forms of HSF1 from E. coli peaked in the same fractions as the protein translated in reticulocyte lysate (data not shown).

FIG. 3.

Properties of HSF1, HSF1[1-430], HSF1[1-382], and HSF1[H179R] extracted from E. coli after limited expression (10 min of IPTG treatment). The concentration range is 1.0 to 2.0 μg ml⁻¹. (A) Whole-cell extracts incubated at the temperatures indicated above the lanes for 20 min were assayed by electrophoretic mobility shift assays (top panels) and analyzed by SDS-PAGE followed by immunoblotting (bottom panels); nonspecific bands served as controls for equal protein loading (asterisks). DNA-binding assays and immunoblots of E. coli extracts are directly comparable. Note that HSF1 was activated in vitro at 37°C, the E. coli growth temperature. B and F, bound and free ³²P-HSE, respectively. (B) Immunoblots showing Superose 6 fractionations of whole-cell extracts incubated at the indicated temperatures for 20 min. A nonspecific band is indicated (asterisks). At the bottom is shown the fractionation of an extract prepared after 50-min IPTG induction; the HSF1 contents of extracts prepared after 10 or 50 min of IPTG induction are also shown (equal amounts of extracted proteins were applied to the gel). (C) Oligomeric properties of HSF1[H179R] purified from E. coli extracts. Left panel: results of SDS-PAGE of Mono Q fraction 9 stained with silver; HSF1[H179R] is indicated (arrow). Lane M, molecular mass markers. Right panel: the same material (∼30 ng) after cross-linking with EGS at the concentrations indicated above the lanes (see Materials and Methods).

The chromatographic behavior of HSF1[H179R], which elutes in the 200-kDa range, is consistent with a HSF1 dimer (or a symmetrical trimer). This mutant was also examined by EGS cross-linking after purification. E. coli extracts were fractionated by three native chromatographic steps (see Materials and Methods), resulting in isolation of nearly homogeneous HSF1[H179R] (68 kDa) (Fig. 3C, left panel) which was active in DNA binding (10) and which cross-linked as dimers, trimers, and a higher oligomer (Fig. 3C, right panel). These results showed that the oligomeric DNA-binding properties caused by the H179R mutation are retained by the pure protein but did not yet solve the subunit composition of this mutant. Furthermore, the purified protein retained the property of being inactivated at 43°C (10).

Since many previous reports have shown that HSE-binding oligomers are extracted from E. coli cells overproducing HSF1, we also expressed higher levels of protein in E. coli. The gel filtration results shown in Fig. 3B (IPTG, 50 min) indicate that, in agreement with those previous data, both oligomers and monomers were present in extracts containing sevenfold more protein, as judged from immunoblots (Fig. 3B, bottom panel).

Intramolecular repression of purified HSF1.

The above results suggest that constituents of reticulocyte lysate do not play an essential role in negative regulation of HSF1 but do not rule out a requirement for an inhibitory protein, which may be conserved in both eukaryotic and prokaryotic expression systems. Therefore, we examined the properties of HSF1 after purification under native conditions (see Materials and Methods). Figure 4A shows that the 68-kDa species recovered after the third step (Mono Q) was nearly homogeneous (>80%), as judged from silver staining, and behaved in gel filtration as a monomer peaking in fractions 21 to 23. Following incubation at 43°C for 5 min, HSF1 formed oligomers eluting in gel filtration fractions 16 and 17 (Fig. 4A, lower panels; note that HSF1 in E. coli whole-cell extracts or in reticulocyte lysate also peaked in these fractions). A total of 30 to 50% of purified HSF1 formed oligomers under this condition, as estimated by densitometric scanning of immunoblots in independent experiments. The corresponding DNA-binding assays (Fig. 4B, left panel) showed that 27 to 68% of purified monomers were activated (see the legend to Fig. 4B). Since longer incubations at 43°C inevitably resulted in some protein loss, HSF1 was supplemented with BSA to determine the temperature profile of activation as a function of time: within the time frame of the experiment (90 min), the highest activity was detected after 5 min at 43°C or 20 min at 37°C; at 30°C, the activation was further delayed and less efficient (Fig. 4C).

FIG. 4.

Inactive HSF1 purified from E. coli extracts. (A) Results of SDS-PAGE showing the purification step on Mono Q (upper left panel); the concentration of HSF1 (a 68-kDa polypeptide [arrow]) was ∼2 μg ml⁻¹ in Mono Q fraction 5, which was used for the following experiments. The protein (40 μl) analyzed by gel filtration was detected by silver staining (upper right panel). Below are shown gel filtrations before and after incubation of the protein at 43°C for 5 min, detected by immunoblotting; prior to this experiment, the protein (15 μl) was diluted threefold in buffer as described in Materials and Methods. (B) Left panel: electrophoretic mobilility shift assays of HSF1 (5 μl) incubated for 5 min at the temperatures indicated above the lanes and pH 7.1. In tests done with ∼220 fmol of protein, the bound HSE was in the range of 20 to 50 fmol after incubation at 43°C for 5 min. Thus, 27 to 68% of monomers (60 to 150 fmol) were activated, assuming binding of one trimer to the single array of three nGAAn modules present in the HSE used in this study. Right panel: activity of HSF1 (5 μl) incubated for 5 min in phosphate buffers at the pHs indicated above the lanes. B and F, bound and free ³²P-HSE, respectively. (C) Electrophoretic mobility shift assays of HSF1 (2.0 μl) incubated at the indicated temperatures and times. The HSF1 stock was supplemented with BSA (1 mg ml⁻¹) prior to aliquoting to prevent protein loss. The 90- and 5-min incubations were started first and last, respectively.

Moreover, the activity was induced by incubations in phosphate buffers of decreasing pH, with a maximum at pH 5.9 (Fig. 4B, right panel). Similar conditions were previously shown to induce HSF activity in crude human and Drosophila cell extracts (28, 52).

The above results show that inactive monomers responsive to temperature and pH changes have been purified to near homogeneity. A 94-kDa species, also present in smaller amounts in the purest fractions, eluted from Mono Q with a different profile and did not seem to form a stoichiometric complex with HSF1 which would elute in earlier gel filtration fractions.

To test if purified monomers are in a repressed conformation, we employed mild proteolysis. For this, the purest fractions were incubated with the broad-range protease chymotrypsin and assayed for DNA-binding activity after addition of the protease inhibitor aprotinin. Figure 5A shows that the DNA-binding activity was induced at increasing protease concentrations (upper panel) and that it correlated with the decrease of the full-size polypeptide and the appearance of proteolytic fragments in the 36- to 45-kDa range (lower panel). The activity was not detected in parallel samples to which aprotinin was also added to inhibit proteolysis (Fig. 5A, lanes 2 and 3). Gel filtration analysis in Fig. 5B showed that the chymotryptic digests contained fragments eluting as apparent oligomers in fractions 19 to 22 (fragments a and c; fragment b and the other minor fragments were not clearly detected), suggesting that fragments excised by proteolysis assembled in HSE-binding oligomers. Similar results were obtained when crude E. coli extracts were incubated with chymotrypsin or protease V8; however, there was no further increase in activity of HSF1 preactivated in crude extracts by either protease (10).

HSF1[1-430], which exhibits temperature-regulated DNA binding in E. coli whole-cell extracts, should show similar properties after purification. The fractionation of E. coli extracts resulted in at least ∼15% homogeneous HSF1[1-430] in Blue Sepharose fraction 7, as judged from silver staining (Fig. 6A, upper left panel), which was used for subsequent tests. Attempts to reach higher homogeneity resulted in severe losses and copurification of an E. coli ³²P-HSE with a similar molecular mass. In gel filtration experiments, HSF1[1-430] eluted as an apparent monomer in fractions 22 and 23, while approximately 50% eluted as a new peak in fractions 16 to 18 after incubation at 43°C for 5 min (Fig. 6A, lower panels), a condition under which DNA binding was activated (Fig. 6A, upper right panel). Finally, incubations in the presence of increasing concentrations of chymotrypsin also resulted in activation of DNA binding (Fig. 6B, upper panel) which correlated with the extent of HSF1[1-430] proteolysis and was inhibited by aprotinin (Fig. 6B, lower panel).

FIG. 6.

Properties of inactive HSF1[1-430] purified from E. coli extracts. (A) Upper left panel: SDS-PAGE gel stained with silver showing fractionation on Blue Sepharose; fraction 7, containing at least ∼15% pure HSF1[1-430] (arrow), was used for the following experiments. Upper right panel: DNA-binding assays of HSF1[1-430] before and after incubation at 43°C for 5 min. Lower panels: gel filtrations before and after incubation at 43°C for 5 min and under conditions described in the legend to Fig. 4. Lane M, molecular mass markers. L.O., load on; fr., fraction. (B) Proteolytic activation of HSF1[1-430]. Upper panel: electrophoretic mobility shift assays after incubation with chymotrypsin at the indicated concentrations or with chymotrypsin and aprotinin as described in the legend to Fig. 5. Lower panel: aliquots of the above samples analyzed by SDS-PAGE and detected by immunoblotting. B and F, bound and free ³²P-HSE, respectively.

DISCUSSION

The mechanism that maintains the inactive state of HSF at physiological temperatures has been previously investigated by mutational analyses using as test systems transfected cells or microinjected Xenopus oocytes. The results of these studies indicated that multiple domains, mainly the conserved carboxy-terminal hydrophobic heptad repeat (HR-C) and the trimerization (HR-A,B) region, are required to maintain both human and Drosophila HSFs in an apparent monomeric form (30, 33, 36, 53). In addition to showing that the monomer-trimer transition and the DNA-binding function are repressed in conjunction, these results led to the proposal that coiled-coil interactions of hydrophobic repeats in the monomer mediate the repression (36). It was also proposed that the inactive monomer is stabilized by other factors, possibly a constitutively expressed 70-kDa HSP (53). Furthermore, gel filtration, sedimentation, and cross-linking studies all indicated that HSF1 extracted from nonshocked cells is a monomer (24, 40, 47), although a fraction of HSF1 was also found associated with HSP70 by use of coprecipitation assays and native polyacrylamide gradient gels (2, 6). However, HSP70 did not exactly fulfill the expected characteristics of a trimerization inhibitor, as it was found associated with HSF1 also in lysates of heat-shocked cells; furthermore, HSP70 overexpression did not inhibit HSF1 activation by heat shock in vivo (37).

The action of an inhibitor in the higher eukaryotes was separately suggested by the observations that HSFs from a broad range of species form oligomers with maximal DNA-binding activity over a wide temperature range in E. coli (7, 11, 22, 26, 27, 49). The properties of mouse HSF1 shown in this study do not contradict these previous observations, since our experiments also show that a substantial fraction of this protein becomes oligomeric during prolonged IPTG treatments of E. coli, most probably due to the elevated concentration.

In this study, we have examined the minimal requirements for repression by comparing the in vitro regulation of mouse HSF1 extracted from E. coli and that of mouse HSF1 translated in rabbit reticulocyte lysate, a system in which mouse and chicken HSF1 were previously shown to exhibit inducible DNA-binding activity (13, 30, 39). Our results show that oligomerization can be repressed in both systems. This is indicated by the observations that HSF1[1-382] and HSF1[H179R] exhibit oligomerization and activity under conditions in which HSF1 and HSF1[1-430] are inactive monomers. Moreover, the properties of HSF1[H179R] and HSF1[1-430] were retained after purification. Thus, the repression requires at least amino acids between positions 382 and 430, containing the HR-C region, and histidine 179 in the HR-A,B region.

To examine the properties of the isolated wild-type monomer, mouse HSF1 was expressed in E. coli and purified under native conditions. Tests with the isolated monomer showed that increased temperature and low pH induce oligomerization and DNA binding and that proteolysis excises fragments that form HSE-binding oligomers. The active species in the mixture of chymotryptic fragments should comprise molecules bearing at least the DNA-binding and oligomerization regions, both located in the amino terminus, and we note that the size range of major fragments in the chymotrypsin and protease V8 (10) digests (36 to 45 kDa) are consistent with the possibility that proteolysis has removed regulatory carboxy-terminal sequences from a fraction of purified monomers.

While chaperone activities of reticulocyte lysate and E. coli may assist the monomeric folding of HSF1, the present results are most consistent with the possibility that elements sufficient for repression, under the conditions we used, reside in the HSF1 molecule, although they do not imply that intramolecular repression is the only mechanism for repression in vivo. Obviously, we cannot rule out that the contaminating 94-kDa species in the purest fractions acts as an inhibitor or that another type of inhibitory molecule was copurified but not detected by silver staining. However, the use of an E. coli system and nearly homogeneous fractions would restrict the range of possible candidates for an inhibitor.

Previous studies by Larson et al. (24) and by Goodson and Sarge (15) showed increases of DNA binding at 43 or 33°C, respectively, in purified protein preparations consisting in one case of renatured human HSF1 after extraction from SDS-polyacrylamide gels and in the other of E. coli recombinant mouse HSF1 expressed as a glutathione S-transferase fusion protein in E. coli. Our results agree with these previous studies in showing that temperature elevations can increase DNA binding in the apparent absence of stoichiometric amounts of other proteins. In addition, our results provide experimental evidence that inactive monomers can be subjected to intramolecular repression.

In both in vitro systems examined in this study, the repression was relieved, although at lower rates, also at temperatures below the heat shock temperature for mammalian cells; i.e., the recombinant HSF1 was not repressed or regulated in vitro as HSF1 was in vivo. It should also be noted that the level of DNA-binding activity induced at 37°C was maximal in the E. coli system but intermediate in the lysate. Differences in HSF1 concentration may account for this discrepancy, but differences in protein folding or modification and the (arbitrary) composition of buffers should also be considered as possible causes. Moreover, an unknown factor(s) might modulate the activation of HSF1 in the reticulocyte lysate, and the modulation of the intrinsic properties of HSF1 by cellular factors might be important also in vivo for setting the induction temperature and the response to stimuli other than heat.

Thus, whereas it remains possible that our choices of recombinant system and experimental conditions were insufficient to reconstitute the physiological regulation of isolated HSF1, reconstitution might require supplementation of HSF1 with other, as-yet-unknown components or a covalent modification of this protein. In the experiments in this study, the elevated temperatures, like other mild denaturants, might simply have caused HSF1 to unfold, thereby mimicking a process which in vivo is controlled by other factors. The observations that human HSF1 can adopt a poised conformation responsive to the stress temperatures of Drosophila, Xenopus, and plant cells (8, 45, 53), although not in all cases tested (8, 18), also suggest that a higher level of regulation is imposed on the foreign HSF (4, 49). Further characterizations of HSFs from normal and heat-stressed cells and improved in vitro reconstitution assays and in vivo assays should help to distinguish between these possibilities and to identify the regulatory factors involved.