Molecular mechanism of attenuation of heat shock transcription factor 1 activity

The heat shock response is a universal transcriptional response to proteotoxic stress orchestrated by heat shock transcription factor Hsf1 in all eukaryotic cells. Despite over 40 years of intense research, the mechanism of HSF1 activity regulation remains poorly understood at a molecular level. In metazoa Hsf1 trimerizes upon heat shock through a leucin-zipper domain and binds to DNA. How Hsf1 is dislodged from DNA and monomerized remained enigmatic. Here, we demonstrate that trimeric Hsf1 is dissociated from DNA in vitro by Hsc70 and DnaJB1. Hsc70 acts at two distinct sites on Hsf1. Hsf1 trimers are monomerized by successive cycles of entropic pulling, unzipping the triple leucine-zipper. This process directly monitors the concentration of Hsc70 and DnaJB1. During heat shock adaptation Hsc70 first binds to the transactivation domain leading to partial attenuation of the response and subsequently, at higher concentrations, Hsc70 removes Hsf1 from DNA to restore the resting state.


Introduction
The heat shock response (HSR) is an ancient transcriptional program, evolved in all organisms to cope with a wide variety of environmental, physiological and developmental stressful conditions that induce an imbalance of protein homeostasis.
This transcriptional program up-regulates hundreds and down-regulates thousands of genes in metazoa 1 . As the HSR was viewed as the paradigm for a homeostatic response, the mechanism of its activation and attenuation was intensively studied in the last four decades but still remains poorly understood at a molecular level 2,3 .
The central role in regulating the HSR and restoring protein homeostasis in eukaryotic cells is fulfilled by the heat shock transcription factor 1 (Hsf1). Through this function Hsf1 is in metazoa at center stage of many physiological and pathophysiological processes like post-embryonic development and aging, cancer and neurodegeneration [4][5][6][7][8] . Metazoan Hsf1 consists of a DNA binding domain (DBD), leucine zipper-like heptad-repeat regions A and B (HR-A/B) functioning as trimerization domain, a regulatory domain (RD), a third heptad repeat region (HR-C) and a transactivation domain (TAD) (Fig. 1A) 9 . Recent structural studies in our laboratory revealed that RD and TAD are mostly unstructured 10 .
In non-stressed mammalian cells Hsf1 is in monomer-dimer equilibrium in the nucleus and the cytosol 10,11 . During stress conditions Hsf1 accumulates in the nucleus and forms oligomers that gain increased affinity for binding to so-called heat shock elements (HSE: nGAAn) arranged in inverted repeats of three and more units [12][13][14] .
Hsf1 HR-C region serves as an intramolecular Hsf1 oligomerization repressor that operates as a temperature sensor, integrating over temperature and time at elevated temperatures by local unfolding and dissociation from HR-A/B 9,10 . The setpoint of this temperature rheostat is dependent on the concentration of Hsf1, allowing nuclear transport and Hsf1 expression to alter the fraction of trimerized Hsf1 at any given temperature 10,15-17 .
Based on accumulated knowledge several Hsf1 activation/attenuation mechanisms have been proposed 18 . The most prominent Hsf1 activity regulation model, the chaperone titration model, assumes that Hsf1 is sequestered and inactivated by molecular chaperones; Hsf1 activation follows the recruitment of the chaperones to stress-denatured proteins. Since Hsp70 was found to co-precipitate with Hsf1, Hsp70 was implicated in repressing Hsf1 in the resting state or during attenuation 19 . Other evidence suggested that Hsp70 is insufficient in metazoa for Hsf1 repression, contesting this model 20 . Since loss of Hsp90 functionality activates the HSR, it was also suggested that Hsp90 chaperones sequester Hsf1 and keep it inactive 21 .
However, to demonstrate Hsp90 and Hsf1 interaction, crosslinking is required 22 , unless an ATPase-deficient Hsp90 variant is used, indicating that Hsf1 interacts with Hsp90 only in the ATP-bound closed conformation 23,24 . In addition, in vitro and in the absence of cochaperones, Hsp90 favors Hsf1 trimerization and DNA binding 10 .
Although genetic evidence suggested an involvement of Hsp90 in HSR regulation also in yeast 25 , more recent ex vivo data suggest that Hsp70 is associated with Hsf1 under non-stress conditions and this interaction is disrupted upon heat shock 26,27 . Whether such a model can be adopted for mammalian cells is not clear since Hsf1 is constitutively trimeric in yeast and does not rely on a monomer-trimer transition for activation 28 . Moreover, the overall degree of sequence identity between yeast and human Hsf1 is just 17% and the proposed binding sites of Hsp70 in yeast Hsf1 are not conserved in human Hsf1.
Several different models have been proposed for HSR attenuation, Hsf1 dissociation from DNA and recycling of Hsf1. Hsp70 and its co-chaperone DnaJB1 have been suggested to bind to Hsf1 within the TAD (aa 425-439) thus attenuating Hsf1 activity by repressing the recruitment of the transcriptional machinery 29 . Hsf1 acetylation in the DBD was proposed to be required for removing Hsf1 from DNA 30 . Thereby the SIRT1 deacetylase plays a main role in delaying the attenuation of the HSR. The interaction between DNA and Hsf1 DBD relies on electrostatic contacts 31 and replacement of Lys80 and/or Lys118 in Hsf1 to glutamate significantly reduced DNA binding 30 . Since the inhibition of proteasomal function also delayed HSR attenuation, it was suggested that, after acetylation-dependent dissociation, Hsf1 trimers are directed to proteasomal degradation 32 . However, inhibition of the proteasome also leads to accumulation of misfolded proteins in the cytosol, eliciting the HSR. For Drosophila Hsf1 it was shown that trimers disassemble spontaneously to monomers at low concentrations 33 . However, such a spontaneous dissociation was not observed for human Hsf1 10 .
In this work we demonstrate using purified components that both Hsc70 and Hsp70 in cooperation with DnaJB1 dissociate trimeric Hsf1 from DNA in the absence of Hsf1 acetylation. Furthermore, we show that during dissociation Hsf1 is monomerized, with most of Hsf1 remaining sequestered in complex with Hsc70. We identify several binding sites for Hsc70 within Hsf1, one of which in the transactivation domain involved in initial attenuation, a second close to the trimerization domain essential for Hsc70mediated monomerization. We provide evidence that Hsc70-mediated monomerization of Hsf1 trimers occurs through stepwise unzipping of the triple leucinezipper of the Hsf1-trimer by entropic pulling. Mutational alteration of the Hsc70 binding sites potentiates expression of a heat shock reporter in HSF1 -/mouse embryonic fibroblasts. Based on these and published data we propose a comprehensive model for a dynamic regulation of Hsf1 activity that closely monitors availability of cellular Hsc70 and Hsp70. Fluorescence polarization plotted versus the Hsf1 trimer concentration. KD = 4.7 ± 1.9 nM (n = 4). C, Hsf1 rapidly switches between different HSE-containing double stranded DNA oligonucleotides. Trimeric Hsf1 was pre-incubated with Alexa 488-labeled HSE-DNA in MgCl2-free buffer. At timepoint 0 (arrow) buffer, 4 mM MgCl2, 2 mM ATP+4 mM MgCl2, control DNA (DNACtrl at 10-fold molar excess over HSE-DNA), and HSE-containing DNA was added as indicated and fluorescence polarization followed over time. Shown is one of three identical experiments. D, Trimeric Hsf1 were bound to HSE-DNA and the indicated components added at timepoint 0. A representative experiment is shown. E, Hsp70 dissociates Hsf1 from DNA with similar rates as Hsc70. Rates were determined as shown in Fig. S1C. F-I, The rate of Hsc70/DnaJB1-mediated dissociation of Hsf1 from HSE-DNA depends on temperature (F), concentration of Hsc70 and DnaJB1 (G), concentration of ATP (H), and concentration of the nucleotide exchange factor Apg2 (HSPA4) (I).

Hsf1 can migrate rapidly between different HSE-containing DNAs
To investigate DNA binding of purified human trimeric Hsf1, we used the previously established fluorescence polarization assay with Alexa488-labeled double stranded DNA oligonucleotides containing 3 inverted HSEs. We first titrated Hsf1 and established the dissociation equilibrium constant KD to ≤ 5 nM (Fig. 1B)

Hsc70 but not Hsp90 can dissociate Hsf1 from DNA
A recent publication proposed an inhibitory role of Hsp90 during the attenuation phase of the HSR 24 . However, in our in-vitro-polarization DNA-binding assay neither human Hsp90α wild type nor its ATPase-deficient variant Hsp90α-E47A, which was shown to bind Hsf1 with higher affinity, had any influence on the change in polarization as compared to the controls (Fig. S1A), indicating that its effect during attenuation phase of the HSR was not achieved through dissociation of Hsf1 from DNA. This result is consistent with earlier findings that Hsp90 promotes Hsf1 trimerization and DNA binding 10 .
In contrast, human Hsc70 in the presence of ATP and its J-domain cochaperone DnaJB1/Hdj1, which targets Hsc70 to client proteins by stimulating Hsc70's ATPase activity, efficiently dissociated Hsf1 from DNA (Fig. 1D). This effect was not observed when any of the three components, Hsc70, DnaJB1 or ATP, was missing, or when Hsc70 wild type was replaced by its ATPase-deficient variant Hsc70-K71M, or its polypeptide binding defective variant Hsc70-V438F, or when DnaJB1 wild type was replaced by a variant (DnaJB1-H32Q,D34N) that is not able to stimulate Hsc70's ATPase activity (Fig. S1B).
When analyzing the shape of the dissociation curve, we observed a short 5 to 10 min delay, during which the dissociation did not follow an exponential function, before the actual exponential dissociation phase started. The data were therefore fitted by a composite function and the rate only represents the exponential phase of the dissociation reaction (Fig. S1C). The dissociation rate was not significantly different whether we used Hsf1 purified from E. coli as trimer and not heat shocked or as monomer and subsequently heat shocked at 42°C for 10 min (Fig. S1D). Also, the heat inducible Hsp70 (HSPA1A/B) dissociated Hsf1 from DNA with similar rates as the constitutive Hsc70 (HSPA8) (Fig. 1E). The reaction was, as expected, temperature dependent and increasing the temperature from 25 to 37°C increased the dissociation rate significantly (Fig. 1F). The kinetics of Hsc70-mediated Hsf1 dissociation from DNA were very similar to the kinetics with which Hsf1-mediated transcription activation and DNA binding of Hsf1 decreased in HeLa cells during recovery after a heat shock 34 .
The dissociation reaction was also strongly dependent on the ATP concentration between 0.05 to 0.5 mM, but not between 0.5 to 2.5 mM (Fig. 1H). In the presence of physiological concentrations of ATP, the life-time of an Hsc70-client protein complex is limited by nucleotide exchange that is accelerated by nucleotide exchange factors 35 . We therefore added the nucleotide exchange factor Apg2 to the dissociation reaction. At very low concentrations, Apg2 accelerated the dissociation reaction, but at higher concentrations it strongly inhibited the reaction and prevented Hsc70-mediated dissociation of Hsf1 from DNA (Fig. 1I). This is similar as Apg2 action in Hsc70-mediated protein disaggregation, where also low concentration of Apg2 stimulate and high concentrations inhibit the reaction 36 . Taken together, Hsc70 and Hsp70 dissociate Hsf1 from its binding sites in promoter DNA in the presence of DnaJB1 and physiological concentrations of ATP in a strongly concentration dependent manner.
This reaction is independent of Hsf1 acetylation in the DBD. . At the timepoints indicated by the arrows samples were taken and separated by blue-native polyacrylamide gel electrophoresis, blotted onto a PVDF membrane and detected with an Hsf1-specific antiserum (B). Lanes 1, purified Hsf1 trimer (T); 2, purified Hsf1 monomer (M); 3-11, samples from the Hsc70/DnaJB1-mediated Hsf1 dissociation reaction (0 to 80 min); 12-15, Dissociation reaction incubated for 80 min missing individual components as indicated. HO, higher order oligomers; T, trimer; D, dimer; M, monomer. C, Quantification of the amounts of Hsf1 species of the blot shown in B and two similar blots as indicated by the brackets to the right; upper bracket, DNA bound timers and higher order oligomers; lower bracket, monomers and Hsc70-bound species. Shown are means ± SD (n = 3).

Hsc70 dissociate Hsf1 from DNA by monomerization of Hsf1 trimers
For the Hsc70-mediated dissociation of Hsf1 from DNA different mode of actions are imaginable. In analogy to Hsp70 action on p53 37,38 , Hsc70 could directly interact with the DBD of Hsf1 to competitively or allosterically remove the DBD from the DNA.
Alternatively, Hsc70 could monomerize Hsf1, which would lead to dissociation from DNA because individual DBDs have only a very low affinity for the HSEs and the high affinity of Hsf1 for heat shock promoters is an avidity effect of three DBDs binding simultaneously. To test the second hypothesis, we followed the dissociation reaction by fluorescence polarization and took samples every 10 min to analyze the oligomeric state of Hsf1 by blue-native polyacrylamide gel electrophoresis (BN-PAGE) and immunoblotting with Hsf1 specific antisera (Fig. 2). In the beginning of the reaction only trimeric and higher order oligomeric Hsf1 was present. In the course of the dissociation reaction the trimer band disappeared and some Hsf1 monomer became visible. Most of Hsf1 exhibited an electrophoretic mobility that is in between the trimeric and the monomeric state, presumably bound to Hsc70. If any of the components were missing, the Hsf1 remained oligomeric. Quantification of the Hsf1 oligomer band and the rest of the Hsf1 species revealed that the oligomer band disappeared within experimental error with the same rate as Hsf1 dissociated from DNA in the fluorescence polarization assay, strongly arguing that Hsc70 dissociates Hsf1 from DNA by monomerization.
Hsc70 also monomerized trimeric Hsf1 with the same rate in the absence of DNA (Fig.   2C).

Hsc70 binds to several sites in monomeric and trimeric Hsf1 and destabilizes the trimerization domain
To elucidate how Hsc70 dissociates Hsf1 trimers, we first wanted to identify the binding site of Hsc70 within Hsf1 using hydrogen exchange mass spectrometry (HX-MS), which is suitable to detect solvent accessibility of amide hydrogens of the peptide backbone and thus conformational changes in proteins and protein-protein interactions 39 , as described in detail previously 10,40 . Briefly, we pre-incubated Hsf1 in the absence or presence of Hsc70 or DnaJB1 or both for 0 or 30 min at 25°C, diluted the sample 1:10 in D2O containing buffer and incubated for 30 or 300 s at 25°C. Subsequently, the exchange reaction was quenched and the samples analyzed by LC-MS, including online digestion of the proteins by immobilized pepsin to localize the incorporated deuterons to specific segments of the protein. Plotting the deuteron incorporation into Hsf1 in the presence of Hsc70 or DnaJB1 minus the deuteron incorporation in the absence of chaperones revealed 5 regions of significant protection that were observed for trimeric as well as for monomeric Hsf1, suggesting 5 potential binding sites (Fig. 3A, Fig. S2). Surprisingly, we did not detect any protection close to amino acids 395-439 previously proposed to harbor the Hsc70 binding site in Hsf1 29 .
Based on an Hsp70 binding site prediction algorithm, originally derived from peptide library scanning data for the E. coli Hsp70 homolog DnaK 41 , two of the segments (200-213 and 442-474) protected from hydrogen exchange by Hsc70 covered sequences that fitted properties of strong Hsp70 binding sites (Fig. 4A, values below -5, dashed line). Close inspection of the sequence of these two segments revealed that both contained two potential Hsc70 binding sites.
To confirm Hsc70 binding to these sequences, we used peptides encompassing the respective sequences labeled with fluorescein and titrated Hsc70 concentration measuring fluorescence polarization. For all four potential binding sites we could determine a KD between 5 and 30 µM (Fig. 4B). To the other protected segments Hsc70 did not have any measurable affinity. The highest affinity Hsc70 binding site was in the TAD, residues 461-471. All sites protected by Hsc70 were also protected by DnaJB1 consistent with the generally accepted mechanism of Hsp70 systems that J-domain proteins bind to the client protein first and target Hsp70 to its clients 35 .
When Hsc70, DnaJB1 and ATP were added to Hsf1 and pre-incubated for 30 min before dilution into D2O buffer and incubation for 300 s, we observed significant deprotection in three segments encompassing the trimerization domain (Fig. 3B, Fig.   S3A), consistent with Hsc70-mediated monomerization. Close inspection of the respective spectra revealed a bimodal distribution of the isotopic peaks, indicative for the coexistence of two subpopulations with different exchange properties. Fitting the two Gaussian distributions to the isotope peak intensities allowed to extract the fraction For relative deuteron incorporation see Fig. S4A. C, Fit of two Gaussian distributions to the peak intensities for peptic peptides 159-168 (left), 169-175 (middle), and 176-189 (right). Representative plots of 3 independent experiments. Original spectra and individual Gaussian distributions are shown in Fig. S4B and C. D, Fraction of high exchanging subpopulation for peptides 159-168, 169-175, and 176-189 under the different indicated pre-incubation conditions: -chaperones, +Hsc70 (70), +DnaJB1 (B1), and +ATP incubated for 0 or 30 min at 25°C. Statistical analysis: ANOVA with Sidak's multiple comparison of different conditions for each of the peptides (n = 3); **, p < 0.01; ***, p < 0.001. E, Relative deuteron incorporation of the low (solid bars) and high (open bars) exchanging subpopulations. ANOVA, Sidak's multiple comparison test; *, p < 0.05; ***, p < 0.001; ****, p < 0.0001. of high and low exchanging subpopulations (Fig. 3C, Fig. S3B and C). In the absence of chaperones, in the presence of ATP, or in the presence of Hsc70, DnaJB1 but without ATP or with ATP but without pre-incubation only a small fraction of the peptides belonged to the high exchanging subpopulation (≤10% for 159-168 and 169-175; ≤17% for segment 176-189). Whereas upon pre-incubation for 30 min at 25°C in the presence of Hsc70, DnaJB1 and ATP the high exchanging subpopulation increased to 20 to 30%. In this case the high exchanging subpopulation exchanged 67 to 91% of the amide protons whereas the low exchanging subpopulation only exchanged 7 to 37% ( Fig. 3E), indicating that in the presence of Hsc70, DnaJB1 and ATP the helices of the leucine-zipper are unfolded.

Binding sites adjacent to HR-B are essential for Hsc70-mediated Hsf1 monomerization
To elucidate whether the Hsc70 binding sites in the TAD are responsible for Hsc70mediated dissociation of Hsf1 from DNA and also to assess any involvement of the previously suggested Hsc70 binding site, we deleted the entire TAD, residues 409-529. Hsf1(1-408) was dissociated from DNA by Hsc70 with similar kinetics as wild type Hsf1 (Fig. 4C), excluding any involvement of sites in the TAD in this process.
Consequently, we deleted the sites close to the HR-B region, residues 202-213. This deletion in Hsf1 completely prevented dissociation of Hsf1 from DNA in the presence of Hsc70, DnaJB1 and ATP, indicating that these residues are essential for Hsc70mediated Hsf1 monomerization (Fig. 4D). Since a typical Hsc70 binding site consists of a core of up to five hydrophobic residues, preferably leucine, flanked by positively charged regions 41 , we exchanged the hydrophobic residues against serine, a small hydrophilic residue disfavoring Hsc70 binding, generating the two Hsf1 variants, Hsf1-I202S,L203S,V205S and Hsf1-I209S,L211S,L213S. Hsc70-mediated dissociation of Hsf1 from DNA was either completely abrogated by these amino acid replacements or strongly inhibited (Fig. 4D). Peptides harboring the same amino acid replacements were also not bound by Hsc70 (Fig. 4B).
Taken together these data indicate that Hsc70-mediated dissociation of Hsf1 from DNA requires binding of Hsc70 to a site close to the trimerization domain. Hsc70 binding to the TAD may serve a different purpose.  Fig. 5A. When we subjected these mixtures to Hsc70-mediated dissociation from DNA, we observed that the rate of dissociation decreased significantly with decreasing Hsf1wt:Hsf1∆(202-213) ratios ( Fig. 5A, Fig. S4B). Assuming that binding of a single Hsc70 is sufficient for Hsf1 monomerization and thus dissociation from DNA at wild type rates and that only the Hsf1∆(202-213) homotrimer cannot be dissociated from DNA, as shown above, the overall rate of dissociation should not change for the different mixing ratios. Only the total amplitude of the dissociation reaction should change, because the Hsf1∆(202-213) homotrimers would remain bound to a fraction of the DNA and thus cause a residual average polarization higher than the polarization of unbound DNA. We simulated this situation and found that the corresponding curves do not fit our experimental data (dashed lines in Fig. S4A). The alternative hypothesis that always three Hsc70 binding sites are necessary would lead to a similar situation, just with smaller amplitudes, also not explaining our experimental data. We concluded that the rate of Hsf1 monomerization depends on the number of Hsc70s bound simultaneously or sequential to different protomers of the Hsf1 trimer. An equation describing this situation fitted our experimental data reasonably well (Fig. S4C). Binding of a single Hsc70 per Hsf1 trimer is able to monomerize the Hsf1 trimer albeit at a very low rate, whereas action of Hsc70 on two protomers of the Hsf1 trimer more efficiently monomerizes Hsf1 trimers, and action of Hsc70 on all three protomers within the Hsf1 trimer is still more efficient in monomerization (Fig. S4D).

Hsc70 monomerizes Hsf1 by entropic pulling
The concept of entropic pulling was proposed for Hsp70 action during import of polypeptides into mitochondria and for protein disaggregation 42 . Briefly, Hsp70 binds to incoming polypeptides close to the membrane. Due to the excluded volume of the bulky Hsp70 the conformational freedom of the polypeptide is limited and thus the entropy low. Movement of the peptide into the mitochondrial matrix increases the distance of the polypeptide bound Hsp70 to the membrane, allowing for more conformational freedom of the polypeptide and thus increases the entropy. Since chemical reaction can be driven by increase in entropy as well as decrease of enthalpy, a force is generated that pulls the polypeptide into the mitochondrial matrix. De Los Rios et al. calculated a pulling force of around 10 to 20 pN that decrease with increasing length of the incoming polypeptide and will reach 0 pN once about 30 residues are imported. To drive further import a new Hsp70 needs to bind to the incoming polypeptide close to the membrane. To test this hypothesis, we moved the Hsc70 binding site away from the HR-B region along the intrinsically disordered regulatory domain. Already when the Hsc70 binding site is 10 residues away from HR-B, Hsc70 dissociated Hsf1 from DNA with a significantly lower rate (Fig. 5B). At a distance of 20 residues Hsc70 was not anymore able to dissociate Hsf1 from the DNA, indicating that monomerization was not anymore possible. These results suggest that Hsc70 monomerizes Hsf1 trimers by entropic pulling.
To substantiate this hypothesis, we inserted a FLAG epitope between HR-B and the Hsc70 binding site or at 10 and 20 residues distance to HR-B. We treated anti-FLAG antibodies with DTT to split them in half (Fig. S5) and added them to DNA bound Hsf1.
Surprisingly, we did not observe any dissociation of Hsf1 (Fig. 5C). This was not due to a failure of the FLAG-antibody halfmers to bind to the FLAG epitope containing Hsf1 trimers as demonstrated by BN-PAGE (Fig. 5D).
We hypothesized that pulling from a single site at the end of the trimerization domain may not be sufficient to unzip the entire domain, since the trimerization domain has a length of 75 residues and the entropic pulling force failed already when Hsc70 bound more than 20 residues away from the leucine-zipper. Close inspection of the HR-A/B region revealed that the sequence contains a large number of hydrophobic residues, as expected for a leucine-zipper, but unexpectedly the C-terminal part of the zipper (HR-B) contains 5 positively charged residues, which favor Hsc70 binding, and not a C, Anti-FLAG antibodies are not able to dissociate HSE-DNA-bound Hsf1. Red lines in the cartoon indicate the Hsc70 binding site; blue lines and arrow head indicate the inserted FLAG epitope DYKDDDDK. Hsf1_i201/211/221, FLAG-epitope inserted after residue 201, 211 or 221. Anti-FLAG antibodies were split in halfmers by incubation with 2 mM DTT (see Fig. S5). D, Anti-FLAG antibody halfmers bound to FLAG-epitope containing Hsf1 variants. Hsf1wt and FLAG-insertion variants were analyzed by blue-native gel electrophoresis in the absence or presence of DTT-treated anti-FLAG antibodies as indicated and Hsf1 detected by immunoblotting; t-bound, FLAG antibody-bound Hsf1 trimers; m-bound, FLAG antibody bound Hsf1 monomer. E, Predicted Hsc70 binding sites in the trimerization domain of Hsf1. Hsp70 score of the trimerization region residues 130-216 of wild type Hsf1 (black); Hsf1-I190S,I194S, magenta; Hsf1-I202S,L203S,V205S, blue; and Hsf1-I209S,L211S,L213S, green. Values below -5 are considered good Hsc70 binding sites. Above the graph is the trimeric homology model of the trimerization domain residues 130-203 with side chains of hydrophobic and charged residues in space filling representation in atom colors with carbon in black except for Ile190 and Ile194 where carbon is shown in green. Above the model is the corresponding sequence. Positively charged residues, blue; negatively charged residues, red. Lines below the sequence indicate two distinct regions with net charge +1 and +5. F, Trimer-to-monomer ratio of freshly purified Hsf1wt and Hsf1-I190S,I194S in the absence of heat shock, determined by gel filtration and BN-Gel (see Fig. S6). Hsf1wt, mean ± SD (n = 3). G, Rate of Hsc70/DnaJB1-mediated dissociation of HSE-DNA bound Hsf1wt and Hsf1-I190S,I194S; mean ± SD (n=3); **, P < 0.01; (paired t-test).
single negatively charged residue, which disfavor Hsc70 binding. Thus, this region of the trimerization domain contains several potential Hsc70 binding sites, as also evident from the Hsc70 binding site prediction (Fig. 5E). To compromise Hsc70 binding in this region is rather difficult, since replacing hydrophobic residues could disturb the leucinezipper and prevent Hsf1 trimerization altogether. We used a model of the trimerization domain kindly provided by A. Bracher 31 which ended at residue 182 and used the iTASSER homology modeling software (https://zhanglab.ccmb.med.umich.edu/I-TASSER/; 43,44 ) to extend the model to residue 203. In this model we discovered two isoleucine residues (I190 and I194) that did not point towards the zipper interface.
Replacing these two isoleucines by serines only moderately reduced the propensity of this region to bind to Hsc70 (red line in Fig. 5E) as compared to the more drastic changes introduced by replacing I202, L203 and V205 (blue line) or I209, L211, and L213 (green line) by serines. Surprisingly, when we purified Hsf1-I190S,I194S we retrieved by gel filtration significantly less trimers and more monomers as compared to Hsf1wt (Fig. 5F), suggesting that the amino acid replacements have destabilized the trimeric state. Nevertheless, Hsc70-mediated monomerization and thus dissociation from DNA occurred at significantly lower rate for Hsf1-I190S,I194S than for Hsf1wt.
These data suggest that binding of Hsc70 to this region contributes to Hsf1 monomerization. Altogether our results suggest that Hsc70 monomerizes Hsf1 by successive entropic pulling unzipping the leucine-zipper step by step.

Amino acid replacements in Hsc70-binding sites potentiate heat shock reporter expression
To test the consequences of our findings in an in-vivo-model system, we stably transfected HSF1 -/mouse embryonic fibroblasts (MEF) with a heat shock reporter expressing mTag-BFP under the control of the HSPA6 promoter. We then transiently transfected these cells with plasmids expressing wild type and mutant Hsf1 and rhGFP in an artificial operon using an IRES element (Fig. 6A) and analyzed the cells by flow cytometry, comparing the median BFP fluorescence level in GFP positive cells (Fig.   6B). Deletion or mutation of the HR-B proximal Hsc70 binding site increased BFP expression levels by up to fourfold and mutation in the Hsc70 binding site in the TAD increased BFP expression roughly twofold (Fig. 6C). In contrast, deletion or mutation of the previously proposed Hsc70 binding site in the TAD did not increase BFP expression. These data clearly demonstrate that binding of Hsc70 to the HR-B proximal Hsc70 binding site is important for attenuation of heat shock gene transcription and binding of Hsc70 to the newly identified Hsc70 binding site in the TAD contributes to attenuation presumably through interference of Hsf1 interaction with the core transcription machinery.

Discussion
In this study we gained several important insights into the regulation of the HSR. We demonstrate that Hsc70 together with its J-domain co-chaperone DnaJB1 removes Hsf1 from heat shock promoter DNA by monomerizing the Hsf1 trimers. Thus, the HSR can be shut off without HSF1 acetylation or degradation and HSF1 can be recycled.
Hsf1 monomerization starts from a Hsc70 binding site C-terminal of the trimerization domain proximal to HR-B and proceeds towards the N-terminus of the trimerization domain through stepwise unzipping the leucine-zipper by entropic pulling (Fig. 7A). We show that starting this repeated binding of Hsc70 at several protomers of the Hsf1 trimer allows for more rapid disassembly. This mechanism makes the biggest contribution to attenuation of heat shock gene transcription in a cell culture model. We also found that binding of Hsc70 to a newly identified site in the TAD contributes to attenuation in the cell culture model, most likely by interfering with the Hsf1-mediated release of stalled RNA polymerase. It is interesting that this binding site has the highest affinity for Hsc70 (KD ca. 5 µM), indicating that increasing concentrations of free Hsc70 during heat shock transcription will on average first bind to this site before binding to the HR-B proximal site.
One might wonder why the affinity of Hsc70 for the critical HR-B proximal site is so low (KD ca. 30 µM). Firstly, it is important to understand that the equilibrium dissociation constant determined for the ADP bound state only insufficiently describes the real affinity of Hsc70 for a binding site. In the presence of ATP and a J-domain protein, Hsc70·ATP associates with high rates with the protein client, then the J-domain protein in synergism with the substrate polypeptide stimulates ATP hydrolysis in Hsc70, leading to transition to the so-called high affinity state with low substrate dissociation rates. This targeting mechanism leads to a non-equilibrium situation that decreases the apparent KD by several orders of magnitude and was coined ultra-affinity 45 . The actual affinity of Hsc70 to this site depends on the local concentration of DnaJB1, which seems to have also some affinity for these sites to increase local concentration and the concentration of the nucleotide exchange factor (Fig. 1I) that allows for ADP release, ATP rebinding, and substrate release. Secondly, the cellular concentration of Hsc70 averaged over 11 different cancer cell lines is around 12-18 µM under nonstress conditions 46 assuming a total cellular protein concentration of 100 -150 mg/ml 47,48 and can reach 14 to 21 µM upon a mild heat shock of 41°C for 4 h 48 . In unstressed cells Hsf1 is in monomer-dimer equilibrium, occasionally trimerizing depending on the local concentration. Trimeric Hsf1 either binds to HSE-promoter DNA driving heat shock gene transcription by releasing paused RNA polymerase or is disassembled immediately by Hsc70/DnaJB1 action (not shown for clarity). Hsc70 binds dynamically to the TAD of DNA-bound Hsf1 trimers attenuating transcriptional activity and bind to its HR-B proximal binding site, disassembling Hsf1 trimers and thus removing it from heat shock promoters. Release of Hsf1 from Hsc70 restarts this cycle. Upon heat shock Hsf1 trimerizes at elevated rates due to its thermosensory function, integrating over temperature and time at elevated temperatures. Simultaneously, Hsc70/DnaJB1-mediated attenuation and disassembly is slowed down, due to binding of Hsc70 to misfolded and aggregated proteins. Both parts of the cycle shift the Hsf1 pool rapidly to the DNA bound active state, accelerating heat shock gene transcription. Elevating Hsc70 concentrations successively attenuate transcriptional activity of Hsf1 and disassemble Hsf1 trimers.
In contrast, human Hsp70 (HSPA1A/B) is barely detectable in most non-cancer cells and reaches only about 0.04 to 0.06 µM upon heat shock 48 , but can reach around 0.7 -10 µM in cancer cells 46 . It is very likely that the HR-B proximal site evolved for such an affinity to Hsc70 to allow for a high enough concentration of Hsc70 in the cell.
If the affinity of this site for Hsc70 were higher, Hsc70 might disassemble Hsf1 already at lower concentrations when the cell still needs more Hsc70 and other stress proteins.
This also explains why the reaction is so exquisitely sensitive to the concentration of Hsc70 and DnaJB1 and why high concentrations of the nucleotide exchange factor Apg2 inhibit the reaction. Apg2 accelerates nucleotide exchange and thus release of Hsc70 from Hsf1. If the first Hsc70 is released from Hsf1 before another Hsc70 can bind to the next Hsc70 binding site that becomes transiently accessible through entropic pulling at the first site, unzipping cannot be efficient. Our data also demonstrate that working on several protomers of the Hsf1 trimer more efficiently disassembles Hsf1 trimers than pulling on a single protomer, providing an additional explanation for the necessity of higher Hsc70 concentrations. A reasonable explanation for this observation is that the entropic pulling force is larger, if several Hsc70 molecules are bound in close proximity to each other to individual protomers of the Hsf1 trimer, increasing local crowding and thereby decreasing the conformational freedom of the HR-B proximal region of the intrinsically disordered regulatory domain.
It is interesting that inserting a FLAG epitope before or after the HR-B proximal Hsc70 binding site and using anti-FLAG antibody halfmers did not lead to disassembly of Hsf1 trimers. This observation contrasts disassembly of clathrin coats that was efficiently performed by replacing the Hsc70 binding site and using FLAG Fab fragments 49 , indicating that a single pull at all three Hsf1 protomers in the Hsf1 trimer is not sufficient for disassembly. It would not be possible to add more FLAG epitopes along the trimerization domain because they would disrupt the triple leucine-zipper.
Conceptionally, disassembly of Hsf1 trimers is more similar to protein translocation through a membrane and to disaggregation. In both cases successive binding events along a polypeptide chain pulls the protein through the membrane pore and out of the aggregate, respectively.
Based on the here presented data and published literature, we propose the following model for the regulation of the HSR (Fig. 7B). In unstressed cells Hsf1 is in a monomerdimer equilibrium occasionally trimerizing in a concentration and Hsp90-dependent manner and bind to HSEs in the genome. Hsc70 may bind to the TAD to attenuate heat shock gene transcription and to the HR-B proximal Hsc70 binding site to disassemble either free Hsf1 trimers before DNA binding (not shown in Fig. 7B)   with delay was fitted to the data using Prism (GraphPad software) (see Fig. S1C).

Molecular mechanism of the attenuation of heat shock factor 1 activity
Szymon W. Kmiecik, Laura Le Breton, and Matthias P. Mayer

Supplemental Information and Figures Supplemental Information
Equation for fitting the dissociation data of mixtures of .
Assuming that a single Hsc70 binding site is sufficient for Hsf1 dissociation and that the number of binding sites does not influence the rate of dissociation the following equation would describe the reaction: Assuming that the number of HR-B proximal Hsc70 binding sites available in the Hsf1 trimer influences the rate by which Hsf1 timers are dissociated, the following equation system has to be used: