Glycogen Synthase Kinase 3 b Negatively Regulates Both DNA-Binding and Transcriptional Activities of Heat Shock Factor 1.

Stress-activation of Heat Shock Factor (HSF1) involves conversion of repressed monomers to DNA-binding homotrimers with increased transcriptional capacity, and results in transcriptional upregulation of the heat shock protein gene family. Cells tightly control the activity of HSF1 through interactions with Hsp90 chaperone complexes and integration into a number of different signaling cascades. A number of studies have shown that HSF1 transcriptional activity is negatively regulated by constitutive phosphorylation in the regulatory domain by glycogen synthase kinase isoforms GSK3 a / b . However, previous studies have not examined the ability of GSK3 to regulate the DNA-binding activity of native HSF1 in vivo under heat shock conditions. Here we show that GSK3 b inhibits both DNA-binding and transcriptional activities of HSF1 in heat shocked cells. Specific inhibition of GSK3 increased the levels of DNA-binding and transcription after heat shock and delayed attenuation of HSF1 during recovery. In contrast, overexpression of GSK3 b resulted in significant reduction in heat-induced HSF1 activities. These results confirm the role of GSK3 b as a negative regulator of HSF1 transcription in cells during heat shock, and demonstrate for the first time that GSK3 b functions to repress DNA-binding. Here we directly examined the in vivo role of GSK3 b on the regulation of endogenous HSF1 under physiological stress conditions. In microinjection experiments with Xenopus oocytes, overexpression of GSK3 b repressed both the DNA-binding and transcriptional activity of endogenous HSF1. In contrast, inhibition of GSK3 with LiCl by overexpression of GBP (GSK3 binding protein) increased DNA-binding and transcription and significantly delayed attenuation of DNA-binding during recovery. These results confirm that GSK3 b acts as a negative regulator of HSF1 transcription in vivo during heat shock, and demonstrate that GSK3 b represses the DNA-binding activity of HSF1.

A number of studies using general kinase inhibitors to examine the potential regulatory role of phosphorylation on the DNA-binding activity of HSF1 have also yielded contradictory information. For example the serine/threonine kinase inhibitor H7 was reported either to inhibit DNA-binding (26) or to have no effect (17,28). Other general kinase inhibitors (GF-X, staurosporine, K252b, KT5720) have no effect on DNAbinding (17,23,28). General phosphatase inhibitors have been shown to have inconsistent effects, either to increase activation (25), delay activation (20), or delay attenuation of DNA-binding (17,28).
Previous studies have clearly demonstrated that amino acids 303, 307, and 363 in the regulatory domain of human HSF1 are constitutively phosphorylated respectively by GSK3, ERK1/2, and PKC, and in vivo analyses of specific mutants have demonstrated that these phosphorylations function to repress the transcriptional activity of HSF1 pXG73/CS2 encoding active GSK3β was generated by PCR cloning from XG73 (kindly provided by D. Kimelman) using PCR primers 5'-cgcggatccatgtcgggaaggccgagaac-3' and 5'-cgcggatcctcaggaggagttggaggcag-3' (33). The PCR product was ligated into the PCR TA3.1 cloning vector (Invitrogen) and the BamH1 insert was further isolated and cloned into pCS2 expression vector. pXG114/CS2 encoding the kinase dead mutant containing a lysine to arginine substitution at position 114 was constructed as above by PCR-cloning from XG114 (provided by D. Kimelman). For expression in oocytes, 20 nl of respective mRNA solutions (2 mg/ml) were injected directly into the cytoplasm. For immediate elevation of nuclear levels of GSK3β, purified recombinant rabbit skeletal muscle GSK3β (New England Biolabs) was injected directly into oocyte nuclei. Cells injected with mRNA were incubated for 12 h at 18°C to allow for translation of expressed protein prior to heat shock. Cells injected with purified enzyme were incubated for 1 h at 18°C prior to heat shock. For stress treatments with LiCl, oocytes were incubated in OR2 containing 10, 25 and 50 mM LiCl for 1 h prior to heat shock in LiCl-free OR2. Heat shock was at 33˚C for 1 h (unless otherwise indicated). For recovery, oocytes were transferred to OR2 at 18°C for the indicated times. In all experiments, a minimum of 25 oocytes were used for each sample.

Transcription and kinase assays
CAT assays were performed using 1 oocyte equivalent of whole cell extract as previously described (35). CAT expression vectors (36) and CAT assay protocols were as previously used in (5). Heat shock treatments were performed 4 h after CAT plasmid injections, then oocytes were incubated 18˚C for 12 hours to allow for CAT expression.
GSK3 activity was assayed as previously described (37). Activity was measured by scintillation counting of label transfer from (γ-32 P)-ATP to the CREB phosphopeptide substrate (New England BioLabs) and expressed as a percentage of activity in uninjected unshocked cells.

GSK 3ß represses HSF1-mediated DNA-binding.
In order to examine whether GSK3β modulates HSF1 in vivo under relevant stress conditions we overexpressed GSK3β and observed the effects on HSF1 activity in Xenopus oocytes. Overexpression was achieved either by microinjection of mRNA into the cytoplasm or by direct injection of purified GSK3β into the nucleus. Increased levels and activity of GSK3β in cells was confirmed by immunoblotting and by specific kinase assays (Fig. 1A). We typically attained a 4 to 5-fold increase in GSK3β protein and activity levels over uninjected control cells (Fig. 1A). By comparison, heat shock for 1 h resulted in a 2-fold increase in endogenous GSK3 activity. In repeated experiments, overexpression of GSK3β consistently resulted in a 5-fold decrease in the amount of HSF1-mediated DNA-binding activity after 1 hr of heat shock ( Fig 1B). Microinjection of purified GSK3β enzyme also resulted in decreased DNA-binding after heat shock ( Fig   1B). Heat-inducible DNA-binding activity was almost completely inhibited after injection of 0.5 U of GSK3β. In order to control for potential non-specific effects on HSF1 due to the microinjection procedure, and to confirm that the effects observed were attributable specifically to alterations in GSK3β activity, we performed similar experiments with an inactive GSK3β mutant lacking kinase activity (kinase dead GSK3β) (38). Expression of similar levels of kinase-dead enzyme did not change total GSK3β kinase activity (Fig. 1A) and had no measurable effect on the amount of DNA-binding by HSF1 after heat shock ( Fig. 1B). Comparison between HSF1 and GSK3 activation during time course of heat shock showed rapid induction of HSE-binding (at 5 min) well before the increase in GSK3 activity which was initially observed after 30 min (Fig. 1C). Control experiments showed that endogenous HSF1 was not activated by the injection procedure itself, and manipulation of GSK3β had no apparent effect on other DNA-binding activities. (Fig.   1D). Therefore, the results of these experiments appear to be specific effects of GSK3β on the DNA-binding activity of HSF1.
In converse experiments, we inhibited endogenous GSK3 by pre-treating cells with LiCl (39) or by overexpression of GBP (GSK3 binding protein), a Xenopus protein which specifically binds to and inactivates GSK3 (32). LiCl and GBP inactivate both GSK3α and β subtypes. Endogenous oocyte GSK3 activity was reduced to less than 10% of controls after treatment with LiCl (25 mM) or after overexpression of GBP (Fig 2A).
Under non-shock conditions, the DNA binding activity of HSF1 was unchanged by GBP expression, but was significantly increased at different time points during heat shock induction ( Fig. 2B). Inhibition by LiCl also resulted in a dose-dependent increase in DNA-binding. DNA-binding was apparently activated in the absence of heat shock at 50 mM LiCl, suggesting that GSK3 may function to repress HSF1 under non-shock conditions.
GSK3β has been hypothesized to facilitate inactivation of HSF1 transcripiton and dispersal of stress-granules following heat stress (15). Therefore we next determined the potential role of GSK3 on the DNA-binding activity of HSF1 during recovery . Inhibition of GSK3 by LiCl or GBP delayed attenuation of DNA-binding activity relative to that GSK3β manipulation caused a general inhibition of transcription was ruled out by controls in which equal expression from the cytomegalovirus (CMV) promoter was observed after each treatment (Fig. 4). It is possible that the decreased transcription observed in oocytes with elevated GSK3β activity was attributable to decreased DNAbinding, however in the experiment in which GSK3β was expressed from microinjected mRNA, DNA-binding was not completely inhibited ( Fig. 1B) whereas transcription was nearly abolished in similarly treated oocytes (Fig. 4). This suggests that GSK3β has dual effects on DNA-binding and transcription. Therefore, these results confirm that GSK3β serves to repress HSF1 mediated transcription under stress conditions in vivo.

Intracellular localization of endogenous GSK3β during heat shock
We have previously shown that HSF1 is a nuclear protein in oocytes before and after heat shock (42). Therefore, direct phosphorylation of HSF1 by GSK3β would require co-localization of these proteins during activation or deactivation. To assess the intracellular distribution of GSK3β before and after heat shock, nuclear and cytoplasmic extracts from manually enucleated cells were analyzed by immunoblotting and kinase assays. GSK3β protein and GSK3 activity were mostly cytoplasmic under non-shock and heat shock conditions and were barely detectable in the nuclei of unshocked cells (Fig.   5). In intact cells, GSK3 activity was elevated by heat shock. The nuclear levels of GSK3β protein were elevated after heat shock, and nuclear kinase activity rose accordingly (Fig 5.). No significant change in GSK3 activity was seen after inhibition of de novo protein synthesis with cycloheximide in both non-shock and heat shock conditions, suggesting that heat shock upregulates endogenous kinase activity. However, the increase in nuclear GSK3 activity could have been caused by a combination of nuclear translocation and activation of the enzyme. In order to demonstrate that the nuclear preparations used in these assays were free of contaminating cytoplasm, we assayed for the presence of IκB and PCNA which are strictly cytoplasmic and nuclear proteins respectively. The cytoplasmic marker IκB was not detected in nuclear extracts and PCNA was not detected in cytoplasmic extracts (Fig. 5).

DISCUSSION
GSK3 is a proline directed serine/threonine kinase that was originally discovered as the major kinase phosphorylating glycogen synthase (43). Various other GSK3 substrates have since been identified, including protein phosphatases, kinases, adhesion molecules, myelin basic protein, and several transcription factors including c-Jun/AP-1, JunD, c-Myb, c-Myc, L-Myc (44). The results shown here clearly demonstrate for the first time that GSK3β negatively regulates the DNA-binding activity of HSF1 in vivo, and confirms that it functions to repress transcription in cells exposed to heat shock. It is interesting that GSK3 phosphorylates c-Jun in the DNA-binding domain and downregulates its DNA-binding activity (45), and that it negatively regulates Myc family members and Myb (46). Therefore, it appears that the inhibitory action of GSK3 may be common to a variety of transcription factors including HSF1.
The role of GSK3β as a negative modulator of HSF1 was supported here by observations that heat-induced DNA-binding was inhibited in cells overexpressing GSK3β, and conversely that DNA-binding was stimulated after inhibition of GSK3. In addition to effects on DNA-binding, our results confirm the role of GSK3 in modulating transcription. HSF1-mediated transcription was either decreased after increasing GSK3β or increased after inhibition of GSK3. In each experiment in which GSK3 was either increased or decreased, we observed consistent and reciprocal effects on HSF1 activities.
The degree to which GSK3β is important for HSF1 regulation in vivo is illustrated by the observation that nuclear enzyme injection almost completely inhibited DNA-binding in response to heat shock. Transcriptional effects on HSF1 by manipulating GSK3 were likely not due to changes in the ability of HSF1 to form trimers and bind DNA because we observed complete transcriptional repression but incomplete repression of DNAbinding with 4-fold increases in GSK3β activity. This leads us to suggest that GSK3 modulates both DNA-binding and transcriptional activities, and this could be through targeting at single or multiple sites on the molecule. The conclusion that GSK3β regulates both activities separately is consistent with uncoupling of DNA-binding and transcription as seen when HSF1 is activated by anti-inflammatory agents such as salicylate and indomethacin (21).
When during the activation-deactivation process does GSK3 exert its influence on HSF1? Previous work suggests that GSK3 might act to constitutively repress HSF1 (13,14,16,30). Consistent with this, we found HSF1 was inhibited in GSK3βoverexpressing cells during early phases of the activation process. We also observed activation of HSF1 DNA-binding under non-shock conditions after inhibiting GSK3 with LiCl, although it is possible that this effect was caused by toxic effects of LiCl. Therefore, it appears that GSK3β represses the DNA-binding and transcriptional activity of HSF1 under normal conditions, and so is important for maintenance of the transcriptionally inactive monomeric conformation. However, several observations lead us to suggest that its primary role is in the later stages of heat shock and attenuation. Our experiments showed that the total level of GSK3 activity increased in response to heat shock well after HSF1 was activated, nuclear levels of GSK increased after 1 hr of heat shock, and inhibition of GSK3 delayed attenuation during recovery. This is consistent with the findings of He et al (15) showing localization of GSK3β to stress granules in vivo and facilitates the disappearance of transcriptionally active HSF1 granules. The current data suggests that GSK3β is required for efficient recovery of HSF1 following resumption of normal conditions. What is the mechanism by which GSK3β-mediated phosphorylation represses HSF1? GSK3β has been shown to phosphorylate human HSF1 at serine 303 (13,14), and so it is likely that it acts directly on HSF1 in oocytes. As with several other GSK3 substrates (47,48), phosphorylation at this site on HSF1 is dependant on hierarchical phosphorylation at an upstream site, (serine 307) by ERK1/2 (13,14). Sequence alignment of human and Xenopus HSF1 (Fig. 6) reveal that the (T286) of frog aligns with serine 303 and matches the consensus phosphorylation site for GSK3β (S/T*XXXS(P)) (47,48). The effects of GSK3β manipulation observed in these experiments could have been through phosphorylation at this site or other site(s) in the Xenopus HSF1 molecule.
It will be interesting to address whether the dual regulation of HSF1 by GSK3 involves targeting of single or multiple sites, and to determine the dynamics of phosphorylations throughout the activation profile.
We hypothesize that GSK3β acts as the terminal kinase of several stress signal transduction pathways that regulate HSF1 activity. GSK3 is known to be the terminal effector kinase of a number of signal transduction pathways. The upstream activators of (49-51). Of these, PKC and PKB are known to be activated by stress, but only PKC has been implicated as a negative regulator of HSF1. Treatment of human erythroleukemia K562 cells with TPA (12-O-tetraecanoylphorbol 13-acetate), a specific activator of PKC, has been shown to enhance heat induced activity, and accelerate attenuation (52) and this effect was abolished by concurrent treatment of cells with a specific PKC inhibitor (53).
In addition, PKC has been shown to directly phosphorylate human HSF1 near the transcriptional activation domain and inhibit transcription (14). It is difficult with the current evidence to determine the relationship between HSF1, PKC and GSK3 at different phases of the stress response. It is interesting that PKC and GSK3 are both activated by stress and are negative regulators of HSF1. Therefore, initial events in the activation of