Heat Shock Transcription Factor 1 Opens Chromatin Structure of Interleukin-6 Promoter to Facilitate Binding of an Activator or a Repressor*

Heat shock transcription factor 1 (HSF1) not only regulates expression of heat shock genes in response to elevated temperature, but is also involved in developmental processes by regulating genes such as cytokine genes. However, we did not know how HSF1 regulates non-heat shock genes. Here, we show that constitutive HSF1 binding to the interleukin (IL)-6 promoter is necessary for its maximal induction by lipopolysaccharide (LPS) stimulation in mouse embryo fibroblasts and peritoneal macrophages. Lack of HSF1 inhibited LPS-induced in vivo binding of an activator NF-κB and a repressor ATF3 to IL-6 promoter. Neither NF-κB nor ATF3 binds to the IL-6 promoter in unstimulated HSF1-null cells even if they were overexpressed. Treatment with histone deacetylase inhibitor or a DNA methylation inhibitor restored LPS-induced IL-6 expression in HSF1-null cells, and histone modification enzymes were recruited on the IL-6 promoter in the presence of HSF1. Consistently, chromatin structure of the IL-6 promoter in the presence of HSF1 was more open than that in its absence. These results indicate that HSF1 partially opens the chromatin structure of the IL-6 promoter for an activator or a repressor to bind to it, and provides a novel mechanism of gene regulation by HSF1.

One of major adaptive responses to high temperature stress in all living organisms is to induce heat shock proteins, which assist protein folding and inhibit protein denaturation (1). This response is regulated mainly at the level of transcription by heat shock transcription factor 1 (HSF1), 2 which can sense an increase in temperature (2,3). The HSF family consists of four members (HSF1-4) in vertebrates, all of which bind to heat shock elements (HSE) (3). In addition to protecting cells from exposure to extreme temperature by inducing Hsp (4,5), HSFs play critical functions in developmental processes such as gamategenesis and neurogenesis (6 -9), in maintenance of the sensory organs (10 -13), and in immune response (14,15), partly by regulating expression of development-related genes such as FGF, LIF, and IL-6 cytokine genes and the p35 gene, an activator of cyclin-dependent kinase 5 as well as Hsp genes (9,11,13,14). Furthermore, HSF1 inhibits expression of tumor necrosis factor-␣ and IL-1␤ by binding directly to the tumor necrosis factor-␣ promoter (16), or by physically interacting with NF-IL6, an activator for the IL-1␤ gene (17). However, it is still unclear how HSF1 regulates expression of non-heat shock genes.
To understand molecular mechanisms of the regulation of cytokine expression by HSF1, we further examined expression of IL-6. Although the effect of HSF1 on IL-6 expression is moderate, HSF1-mediated IL-6 expression may be involved in various aspects of inflammatory and immune response such as antibody production (14). Here, we found a novel function of HSF1 that partially opens the chromatin structure of the IL-6 promoter for an activator or a repressor to bind to it in unstressed conditions.
Northern Blot Analysis and Reverse Transcriptase-PCR-Total RNA was isolated from MEF or tissues using TRIzol (Invitrogen) and Northern blot analysis were performed as described previously (20) using cDNA probes for mouse IL-6, Hsp70-1, and ␤-actin (14). Levels of mRNAs were estimated using the NIH Image program.
Western Blot Analysis and Gel Shift Assay-Cell extracts were prepared from MEF cells in Nonidet P-40 lysis buffer (150 mM NaCl, 1% Nonidet P-40, 50 mM Tris (pH 8.0), 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 mM dithiothreitol), and centrifuged at 12,000 ϫ g for 10 min. Equal amounts of soluble proteins were loaded on SDS-PAGE, and transferred onto nitrocellulose membranes. The cytoplasmic and nuclear proteins were isolated as described previously (21). The membranes were blotted with rabbit antiserum for mouse HSF1 (␣-mHSF1g) (22) and human HSF1   . HSF1 binding to IL-6 promoter is constitutive and nearly constant during heat shock. A, ChIP-enriched DNAs using anti-HSF1 antibody were prepared from wild-type and HSF1 Ϫ/Ϫ cells treated without (Ϫ) or with heat shock at 42°C for 1 h (HS) or LPS for 4 h (LPS). DNA fragments of IL-6 promoter (HSE2 fragment, Ϫ827 to Ϫ565; HSE3 fragment, Ϫ230 to ϩ31) and Hsp70 -1 promoter (Ϫ272 to ϩ47) were amplified by PCR. B, time-dependent profile of HSF1 binding to IL-6 or Hsp70 promoter. ChIP analysis was performed using wild-type cells treated with heat shock at 42°C for the indicated periods. C, apparent dissociation constant of HSF1 was determined for an ideal HSE (HSE81) or an HSE in the IL-6 promoter (HSE2). Saturating binding was performed as described under "Materials and Methods." The amount of labeled probe was 0, 0.125, 0.25, 0.5, 1, 2, 4, and 8 nM. The apparent K d was calculated from the slope of the best-fit line for the Scatchard plot.
(␣-hHSF1e) (we generated it against recombinant hHSF1 fused to glutathione S-transferase), rabbit polyclonal IgG for ATF-3 (C-19, Santa Cruz) or for NF-B p65 (C-20, Santa Cruz), or mouse monoclonal IgG for Hsp70 (W27, Santa Cruz) or GFP (Nacalai Tesque, Japan). Gel filtration analysis was performed as was described previously (21). Whole cell extracts were prepared in buffer C (14) from MEF cells treated with heat shock or cells infected with adenovirus expressing mutant HSF1, and subjected to gel shift assay. A probe used was an ideal HSE oligonucleotide (14).
Saturating binding analysis was done as described previously (23), except that the concentration of DNA probes were varied from 0 to 8 nM and the concentration of the protein was kept constant (10 and 25 g of whole cell extracts from heat-shocked HeLa cells per reaction for HSE81 and HSE2, respectively). The amount of bound and free DNA in saturating binding was quantified by image analysis. The K d for HSE81 and HSE2 was determined from Scatchard analysis and the equation of the best-fit line with the formula K d ϭ Ϫ1/slope. Viral Infection-To generate adenovirus vectors, cDNAs for wild-type human HSF1 and mutants HSF1(R71G) and HSF1(R71A) containing Kozak consensus sequence were amplified by PCR using pZeo-hHSF1, pZeo-hHSF1R71G, and pZeo-hHSF1R71A (18) as templates using primers: hHSF1-5Ј, 5Ј-GGCGGATCCGCCACCATGGATCTGCCCGTGGGC-3Ј and hHSF1-3Ј, 5Ј-CCGCTCGAGCGGCTAGGAGACAGTG-GGGTCCTTGG-3Ј (underlined sequences indicate KpnI and XhoI sites, respectively). Amplified cDNAs were inserted into pShuttle-CMV vector (Stratagene) at KpnI and XhoI sites. Viral DNAs were generated according to the manufacturer's instructions for the AdEasy adenoviral vector system (Stratagene) and viral particles were enriched as described previously (22,24). pZeo-hHSF1⌬AB was generated as described previously (18) using internal primers: HSF1⌬AB-3, 5Ј-CAGGAGTGCATG-GACCGGCAGTTCTCCCTG-3Ј, and HSF1⌬AB-4, 5Ј-CAG-GGAGAACTGCCGGTCCATGCACTCCTG-3Ј. Because an amino acid in the oligomerization domain of HSF4 is mutated (arginine at amino acid 175 was substituted to proline) (25), we generated an expression vector for the corresponding HSF1 mutant, pZeo-hHSF1R176P using internal primers: hHSF1R176P-5, 5Ј-TGGCCAGCCTTCCGCAGAAGCAT-GCCCA-3Ј, and hHSF1R176P-3, 5Ј-TGGGCATGCTTCT-GCGGAAGGCTGGCCA-3Ј. Adenoviral vectors were generated as described above. . Proteins in fractions were subjected to Western blotting using HSF1 antibody (␣hHSF1e). The predicted elution positions of monomeric and trimeric forms of HSF1 are indicated at the bottom. Elution profiles of endogenous HSF1 in control, heat-shocked, and LPS-treated wild-type cells are shown (lower). HSF1 trimer is eluted in fractions 17 to 19. C, cell extracts were isolated from control (C) and heat-shocked (HS, 42°C for 1 h) wild-type and HSF1 Ϫ/Ϫ cells that were infected with adenoviruses expressing the indicated HSF1 mutants, and were subjected to gel shift assay (upper). The same extracts were subjected to Western blotting using antibodies for HSF1 (␣hHSF1e plus ␣mHSF1g) and Hsp70 (lower). Constitutive HSE binding activity was detected when autoradiography was performed for longer periods (upper right). D, ChIP analysis was performed using GFP expressing wild-type cells, and HSF1 Ϫ/Ϫ cells infected with adenoviruses expressing GFP, HSF1, and the indicated HSF1 mutants. DNA fragments of IL-6 promoter (HSE2) and Hsp70 -1 promoter were amplified by PCR. E, wild-type and HSF1 Ϫ/Ϫ cells were infected with adenovirus expressing GFP, HSF1, or a mutant HSF1 (R176P or ⌬AB), and then incubated with LPS for 6 h. Northern blot analysis was performed, and levels of IL-6 mRNA compared with its basal levels in HSF1 Ϫ/Ϫ cells are shown (n ϭ 3). Stars indicate these are significant (p Ͻ 0.05).
Restriction Enzyme Accessibility-Experiments were performed essentially as described previously (26). Briefly, cell nuclei isolated from 2 ϫ 10 7 MEF cells were incubated with AflII (20 units) for 15 min at 37°C. Then genomic DNA was isolated from the nuclei using DNeasy Tissue Kit (Qiagen). The purified DNA (5 g) was digested completely with XbaI and SpeI (20 units each) at 37°C overnight, and 1.8-kb (Ϫ1109 to ϩ626) and 0.9-kb (Ϫ1109 to Ϫ262) DNA fragments were amplified by 30 cycles of PCR using LA Taq polymerase. Primers used were: IL-6g1, 5Ј-GACCCAGCCTAGAAGACTTGAGC-3Ј; IL-6g3, 5Ј-GCAG-TCACATTCTGTATCCTTCCAGACAGG-3Ј; and IL-6g4, 5Ј-CTTTAAAAGTGACTCAGCACTTGAGC-3Ј. The DNA fragments were subjected to Southern blot analysis, and signals were quantified by image analysis. The percentage of cleavage was calculated as described in the legend to Fig. 7.
Statistical Analysis-Significant values were determined by analyzing data with the Mann-Whitney's U test using Stat View version 4.5J for Macintosh (Abacus Concepts, CA). A level of p Ͻ 0.05 as considered significant.

HSF1 Is Required for Maximal Induction of IL-6 Expression-
We showed previously that HSF1 is required for IL-6 to be maximally induced in cultured spleen cells and peritoneal mac-

. HSF1 is required for ATF3 binding to IL-6 promoter in vivo.
A, wild-type and HSF1 Ϫ/Ϫ cells were infected with Ad-ATF3 and Ad-GFP, and protein levels were examined by Western blot analysis (right). Cells infected with each adenovirus were treated with 1 g/ml LPS for 6 h. Northern blot analysis was performed (left). B, wild-type and HSF1 Ϫ/Ϫ cells infected with each adenovirus were treated with LPS for the indicated periods. Time-dependent profiles of LPS-induced IL-6 mRNA levels are shown compared with its control level in wild-type cells (n ϭ 3). C, wild-type and ATF3 Ϫ/Ϫ cells were untreated (C) or treated with heat shock (HS) or LPS for 6 h. Northern blot analysis was performed, and levels of IL-6 mRNA compared with its basal levels in wild-type cells are shown (n ϭ 3) (star, p Ͻ 0.05). D, cells were treated with LPS, and ChIP analysis was performed using anti-ATF3 antibody. DNA fragments of IL-6 (Ϫ230 to ϩ31) were amplified by PCR. E, ChIP analysis using cells infected with Ad-GFP, Ad-ATF3, or Ad-HSF1 for 48 h (upper). The cytoplasmic (C) and nuclear (N) proteins were subjected to Western blot analysis by using antibody for ATF3, GFP, and ␤-actin (lower). NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45 rophages in response to LPS treatment (14). Here, we examined IL-6 expression in primary cultures of MEF because NF-Bregulated genes are induced in response to LPS stimulation (27). Consistent with a previous report, LPS-induced levels of IL-6 mRNA and secreted IL-6 levels in media were much lower in HSF1-null MEF and macrophage cells than those in wildtype cells, and heat shock did not induce IL-6 expression (Fig. 1,  A and B).

IL-6 Regulation by HSF1
When cells were treated with LPS, IL-6 expression is induced mainly by NF-B, and then suppressed by ATF3, a member of the ATF/CREB family of transcription factors (28,29). In addition to NF-B, many transcription factors such as NF-IL6, AP-1, IRF-1, and ATF-2 enhance IL-6 expression (30 -34) (see Fig. 7A). Previous work suggests that HSF1 might directly activate specific transcription factors such as NF-IL6 (17). However, we found that phorbol 12-myristate 13-acetate and serum stimulations moderately induced IL-6 expression and interferon-␥ treatment accelerates LPS-induced expression, whereas IL-6 expression is less induced by all of these stimuli in HSF1null cells (Fig. 1C). Furthermore, we found that constitutive and LPS-induced IL-6 mRNA levels were restored when HSF1 was re-expressed into HSF1-null cells, whereas these were not restored at all when HSF1(R71G) or HSF1(R71A), in which a DNA contact site arginine was substituted to glycine or alanine (18), was re-expressed (Fig. 1D). These results demonstrate that DNA binding of HSF1 is necessary to induce IL-6 expression at a maximal level, and excluded the possibility that HSF1 directly affects some other transcription factors.
HSF1 Binds to IL-6 Promoter on Unstressed Conditions-Comparison of mouse and human IL-6 promoter sequences revealed conserved three HSE-like sequences within Ϫ1,000 bp from transcription start sites (supplemental Fig. S1) (14). ChIP analysis showed that binding of HSF1 was detected in unstressed cells by PCR amplification of an HSE2 fragment (Ϫ827 to Ϫ565) containing overlapping HSE-like HSE1 and HSE2 sequences, but not of an HSE3 fragment (Ϫ230 to ϩ31) containing an HSE-like HSE3 sequence and major regulatory elements ( Fig. 2A). This result strongly suggests that the HSE2 (overlapping with HSE1), which is located in a uniquely distal position among other regulatory elements, is an HSF1-binding site. Even though HSF1 directly binds to the IL-6 promoter, heat shock did not induce IL-6 expression (Fig. 1A). Consistently, the level of HSF1 binding to IL-6 promoter was nearly constant during heat shock although the level of HSF1 binding to the Hsp70 promoter increased significantly (Fig. 2, A and B). Therefore, it is unlikely that HSF1 recruits transcriptional initiation complexes on the IL-6 promoter (2).
To understand the molecular basis of unique interactions between HSF1 and the IL-6 promoter, we performed saturation binding to determine apparent dissociation constants (K d ) (23). HSF1 binding to an HSE2 probe was saturated at a much lower concentration of the probe than that to an ideal probe, and Scatchard analysis revealed that the K d value of HSF1 binding to an HSE2 probe (21.3 nM) was 7-fold higher than that to an ideal HSE probe (3.3 nM) (Fig. 2C).
HSF1 is known to exist as a monomer that cannot bind to DNA in unstressed cells. In response to heat shock, it is converted to a DNA-binding trimer that activates heat shock genes (2,3). However, recent loss-of-function analysis of HSF1 revealed that HSF1 regulates target gene expression in unstressed cells (35,36) and during development (11,13). To determine whether an HSF1 monomer could play any role or not, we generated a deleted HSF1 (HSF1⌬AB) that lacks the oligomerization domain, and point mutant HSF1(R176P) in which arginine at amino acid 176 in the oligomerization domain was mutated to proline (Fig. 3A). Wild-type HSF1 forms a DNA binding trimer when it was overexpressed in cells, whereas HSF1⌬AB was unable to bind to DNA in vitro as it stayed a monomer (Fig. 3, B and C). Overexpressed HSF1(R176P) stayed mostly a monomer in unstressed cells and was unable to bind to DNA in vitro even though it formed a trimer on heat shock conditions (Fig. 3, B and C). We found that the two HSF1 mutants neither bind to the IL-6 promoter in vivo (Fig. 3D), nor restore IL-6 expression in HSF1-null cells (

. HSF1 is required for NF-B binding to IL-6 promoter in vivo.
A, cells were treated with LPS, and ChIP analysis was performed using anti-p65 antibody (upper). The cytoplasmic and nuclear proteins were subjected to Western blot analysis using antibody for p65, GFP, and ␤-actin (lower). B, ChIP analysis was performed using cells infected with Ad-GFP, Ad-p65, or Ad-HSF1 for 48 h (upper). The cytoplasmic and nuclear proteins were subjected to Western blot analysis (lower). 3E). These results strongly suggest that an HSF1 monomer cannot bind to the IL-6 promoter in vivo. A little amount of HSF1 trimer that exists in unstressed cells (Fig. 3, B and C) may bind to the IL-6 promoter in vivo in unstressed cells.
HSF1 Is Required for ATF3 and NF-B to Bind to IL-6 Promoter in Unstimulated Cells-To understand interplay of HSF1 and a transcriptional repressor or an activator on the IL-6 promoter, we examined whether HSF1 is required for ATF3 to suppress LPS-induced IL-6 expression. Surprisingly, overexpression of ATF3 into wild-type cells suppressed LPS induction of the IL-6 mRNA level, whereas its overexpression into HSF1null cells had no effect on the LPS-induced level (Fig. 4, A and  B). The partial inhibitory effect of ATF3 in wild-type cells was expected because loss of ATF3 function also led to a moderate increase of IL-6 expression (Fig. 4C). ChIP analysis identified that constitutive and LPS-induced binding of ATF3 on the IL-6 promoter was much less in HSF1-null cells compared with that in wild-type cells (Fig. 4D). Remarkably, ATF3 never bound to the IL-6 promoter in vivo in unstimulated HSF1-null cells even though ATF3 was ectopically overexpressed in the nucleus (Fig.  4E). These results demonstrate that HSF1 is required for ATF3 to bind to the IL-6 promoter in vivo in unstressed cells, and it promotes LPS-induced ATF3 binding to the promoter.
To examine whether HSF1 is required for binding of only a negative regulator ATF3 or not, we analyzed p65, a component of NF-B that is a major transcriptional activator for LPS-induced gene expression. We found that LPS-induced in vivo binding of p65 was markedly low in HSF1-null cells although most p65 translocated into the nucleus after LPS treatment (Fig.  5A). Furthermore, p65 never bound to the IL-6 promoter in unstimulated HSF1-null cells even though p65 was overexpressed in the nucleus (Fig. 5B). These results clearly showed that HSF1 promotes binding of both an activator and a repressor to IL-6 promoter.
HSF1 Affects Modification of Histone and DNA-As the above results suggested that HSF1 may regulate chromatin structure of the IL-6 promoter on unstressed cells, we treated cells with tricostatin A, a histone deacetylase inhibitor, or 5-aza-2Ј-deoxycytidine, a DNA methylation inhibitor, and found that the inhibitors elevated constitutive and LPS-induced IL-6 mRNA levels in HSF1-null cells to the levels in wild-type cells (Fig. 6A). These treatments did not alter mRNA levels of ␤-actin as well as Hsp70, whose promoter is opening in the absence of heat shock (37). The result indicated that the inhibitors mimic the effects of HSF1, implying the function of HSF1 is for chromatin remodeling. Furthermore, ChIP analysis showed that histone H3 was more acetylated in the presence of HSF1, which is associated with increased recruitment of CBP histone acetylase and BRG1, an ATPase subunit of SWI/SNF nucleosome remodeling complex (Fig. 6B). As HSF1 physically interacts with BRG1 (38) and a plant HSF interacts with CBP (39), our results suggest that HSF1 could recruit CBP and SWI/ SNF complex on IL-6 promoter directly or indirectly in the absence of any stimulus.
HSF1 Promotes Chromatin Opening of IL-6 Promoter-We next examined remodeling of IL-6 promoter-encompassing nucleosomes by HSF1 using the restriction enzyme accessibility assay. Isolated nuclei were treated with AflII restriction enzyme that can specifically digest a site located between NF-B and ATF3 binding sites (Fig. 7A). We found that the IL-6 promoter was not cleaved at all by treating with 20 units of AflII for 15 min in unstimulated HSF1-null MEF cells, although the promoter was efficiently cleaved in unstimulated wild-type cells (Fig. 7B). Cleavage rates of IL-6 promoter in both wild-type and HSF1-null cells increased following LPS stimulation, but the IL-6 promoter in wild-type cells was more efficiently cleaved than that in HSF1-null cells at any time points after LPS treatment (Fig. 7C). IL-6 promoter in macrophages treated with or without LPS was also cleaved more efficiently in the presence of HSF1 (Fig. 7D). Heat shock treatment did not affect the cleavage rate by AflII (Fig. 7E). These results demonstrate that HSF1 is necessary to partially open the chromatin structure under basal conditions and also enhances opening its chromatin structure in response to LPS stimulation, and are consistent with HSF1-dependent accessibility of ATF3 and NF-B on the IL-6 promoter (Figs. 4 and 5).

DISCUSSION
A unique feature of HSF1 regulation on IL-6 expression is that HSF1 facilitates binding of ATF3 and NF-B on the IL-6 promoter in some cells such as MEF cells. Even in the absence of stress, HSF1 partially opens the chromatin structure of the IL-6 promoter, and this activity is also important for a repressor ATF3 and an activator NF-B to bind to the promoter when cells are stimulated with LPS. HSF1 may facilitate LPS-induced chromatin opening by unknown factors. This novel function of HSF1 explains why HSF1 is required for IL-6 to be maximally induced in response to LPS treatment in MEF cells, cultured spleen cells, and peritoneal macrophages (14).
Previous work showed that HSF1 is required for targeting histone acetylation and recruiting chromatin remodeling com-plex to heat shock genes in response to heat shock (40,41), and histone acetylation precedes chromatin opening of the yeast Hsp82 gene (42). However, there was no evidence that HSF1 can induce histone acetylation and chromatin remodeling in unstressed cells. Our results revealed recruitment of histone acetylase and nucleosome remodeling complexes on the nonheat shock IL-6 gene by HSF1 in unstressed cells, which is required for the IL-6 gene to be fully activated or repressed.
As was showed in IL-6 promoters, there are many HSE-like sequences that could be bound by HSFs. In fact, in vivo ChIP analysis showed many non-heat shock genes that are bound by HSF1 (43). HSF is localized at specific loci on chromosomes in unstressed Drosophila salivary glands (44,45), and several development-related genes are under control of HSFs (9,11,13,14). Our observations suggest that HSF1 might play a general role in maintaining the nucleosome-free structure over the transcription start site of constitutively expressed genes.
We do not know the molecular basis explaining why HSF1 can play a role in unstressed cells, as HSF1 mostly stays a monomer that cannot bind to DNA in unstressed conditions (2,3). Remarkably, the level of HSF1 binding to the IL-6 promoter was nearly constant during heat shock although the level of HSF1 binding to the Hsp70 promoter increased significantly (Fig. 2B). This result is unexpected because less than several percent of . HSF1 is required for chromatin opening of IL-6 promoter in unstimulated cells. A, schematic representation of IL-6 promoter. Cis-elements for various transcription factors indicated are shown as boxes, and the restriction site for each enzyme is shown. PCR primers are indicated as g1, g3, and g4, and a probe for Southern blotting is indicated. Primers g1 and g3 only applied uncleaved product (U). Primers g1 and g4 were applied to both uncleaved and cleaved products (T). The percentage of cleavage was calculated by T-U divided by T. B, nuclei from wild-type (ϩ/ϩ) or HSF1-null (Ϫ/Ϫ) MEF cells were subjected to restriction enzyme accessibility using 20 units of AflII for the indicated periods (upper), or using the indicated units for 15 min (lower). C, nuclei from MEF cells treated with LPS for the indicated periods were subjected to restriction enzyme accessibility using AflII (20 units, 15 min). Time dependent changes of ratios of AflII cleavage during LPS treatment were calculated as described above and shown (n ϭ 3). D, nuclei from wild-type (ϩ/ϩ) and HSF1-null (Ϫ/Ϫ) peritoneal macrophages were subjected to restriction enzyme accessibility using AflII (20 units) for the indicated periods (upper). AflII cleavage (20 units, 30 min) during LPS treatment was shown (lower). E, nuclei isolated from wild-type MEF cells incubated with heat shock at 42°C for 1 h or LPS for 2 h were isolated, and subjected to AflII accessibility (20 units, 30 min).
HSF1 existed as a trimer in unstressed MEF cells and nearly 100% of it existed as a trimer in heat-shocked cells (Fig. 3B). In unstressed cells, does an abundant HSF1 monomer bind to the IL-6 promoter? Our analysis strongly suggests that an HSF1 monomer cannot bind to the IL-6 promoter in vivo (Fig. 3). Therefore, binding of a little trimeric HSF1 to its HSE2 sequence could saturate at a much lower concentration as was suggested in Fig. 2C, or binding of an HSF1 trimer to IL-6 promoter might be stabilized by unknown factors in unstressed cells. The IL-6 gene possessing such a unique feature might represent genes to which HSF1 binds constitutively and constantly, irrespective of heat shock.