CHIP activates HSF1 and confers protection against apoptosis and cellular stress

CHIP increases Hsp70 expression by activation of transcription factor HSF1

We examined the cellular functions of one such co-chaperone, CHIP, utilizing an adenoviral approach to overexpress CHIP in COS7 cells, which express CHIP endogenously at low but detectable levels. When protein extracts were analyzed by Coomassie Blue staining, we noted that overexpression of CHIP elevated the levels of a highly inducible 70 kDa protein (Figure 1A). Based on the size and abundance of this protein, we entertained the possibility that CHIP could regulate the expression of stress-responsive chaperones. Indeed, western blot analysis demonstrated that several chaperones were regulated by CHIP, with the effects on Hsp70 being most marked (Figure 1B). CHIP’s induction of Hsp70 was comparable to and not additive with the effects of heat shock (Figure 1C), making it likely that CHIP and heat shock activate Hsp70 through the same pathway. This unique observation of co-chaperone-dependent Hsp70 expression led us to consider its mechanisms and functional consequences in more detail.

A trivial explanation is that overexpression of CHIP increases the concentration of misfolded proteins, which in turn induce the stress response; however, this was not observed in metabolic labeling studies (Meacham et al., 2001), neither has general inhibition of proteasome function been detected (Jiang et al., 2001). To address the mechanism of Hsp70 regulation by CHIP, we performed reporter gene assays using the Hsp70 promoter, as Hsp70 is predominantly regulated at the transcriptional level by nuclear proteins such as HSF1 and STAT family members (Stephanou et al., 1999; Pirkkala et al., 2001). CHIP expression potently activated Hsp70 promoter activity (Figure 1D). The activation by CHIP was comparable to that induced by heat shock and was abolished in a promoter fragment that lacked the HSE, indicating that this effect is dependent on HSE-binding nuclear protein(s). We addressed the role of the Hsp70 HSE in mediating the transcriptional effects of CHIP more directly by gel mobility shift analysis. Specific binding of nuclear proteins to the radiolabeled HSE was enhanced by CHIP overexpression (Figure 1E). This binding could be competed by an excess of unlabeled HSE but not by a non-specific sequence. The identity of this specific binding activity as HSF1 was verified by mobility supershift with an antibody recognizing HSF1. Finally, we confirmed that HSF1 was necessary for the induction of Hsp70 by CHIP in fibroblasts from mice deficient in HSF1 (McMillan et al., 1998; Xiao et al., 1999). CHIP upregulated Hsp70 appropriately in wild-type fibroblasts, but not in cells lacking HSF1 (Figure 1F). In fact, modest downregulation of basal Hsp70 levels by CHIP was observed in HSF1-null cells, which may imply that CHIP has effects on more than one signaling pathway that regulates Hsp70 expression, although HSF1-dependent regulation is clearly the dominant pathway and probably the only positive regulatory pathway affected by CHIP. This may reflect an activity of CHIP on other signaling pathways that also affect Hsp70 expression but that are only obvious when the effects of HSF1 are removed; such an interpretation is consistent with the results of Figure 1D as well.

Evidence from a variety of systems indicates that both Hsp70 and Hsp90 can repress the activation of HSF1, as can other co-chaperones, making the activation of HSF1 by CHIP an exceptional observation. The induction of Hsp70 by CHIP is specific, as BiP levels (which are inducible by activators of the endoplasmic reticulum stress response such as tunicamycin) were unaffected by CHIP (Figure 2A). Importantly, the inducibility of Hsp70 was completely abolished by a mutant form of CHIP (CHIP K30A). This mutant is expressed at similar levels to wild-type protein and differs only in a single amino acid residue in the TPR domain that abolishes interactions with Hsp70 (Scheufler et al., 2000); thus, non-specific activation of the stress response by protein overexpression can be excluded. To further test the specificity of this effect, we examined a panel of co-chaperones for their ability to increase Hsp70 expression and promoter activity. In comparison with other co-chaperones (including protein phosphatase 5, FKBP59 and HOP, each of which contains TPR repeats), only CHIP increased Hsp70 expression (Figure 2B), and no co-chaperone other than CHIP significantly activated the Hsp70 promoter (Figure 2C). The specificity of this effect raised the question of which features of CHIP contribute to HSF1 activation. To address this, we tested the ability of CHIP mutants to increase Hsp70 expression and to activate the Hsp70 promoter in transient transfection assays. CHIP K30A is ineffective in increasing Hsp70 expression or promoter activity, whereas the activity of CHIP D253N/R254G (which contains mutations in the U-box that abolish ubiquitin ligase activity (Xu et al., 2002)) was equivalent to that of wild-type CHIP (Figure 2D). Thus, the activation of HSF1 by CHIP is dependent on CHIP’s ability to interact with Hsp70 and/or Hsp90, whereas the ubiquitin ligase activity may not be as important, at least under these conditions.

Assembly of a CHIP-dependent activated HSF1 complex

One of the hallmarks of HSF1 activation is retarded mobility due to trimerization. Using a non-denaturing gel separation technique that allows discrimination of monomeric, dimeric, Hsp70-bound and trimeric HSF1 complexes (Zou et al., 1998; Liu and Thiele, 1999; Guo et al., 2001), we found that increasing levels of CHIP caused rapid and nearly quantitative accumulation of active, trimeric HSF1 (Figure 3A). Interestingly, accumulation of the activated form of HSF1 coincided with the earliest detectable expression of CHIP after adenoviral infection and persisted for at least 48 h, even though other ongoing stimuli for HSF1 activation (such as heat shock) have a self-terminating effect (in our experiments, within 4 h), suggesting that CHIP-induced trimeric HSF1 complexes may be resistant to attenuation. CHIP had no effect on the stability of HSF1, nor did we detect ubiquitylation of HSF1 by CHIP, but labeling studies did indicate that HSF1 was phosphorylated coincident with CHIP-induced trimerization (data not shown), as is suggested by the slightly retarded mobility of HSF1 we consistently noted by western blotting after denaturing gel electrophoresis. Several studies have demonstrated that both Hsp70 and Hsp90 can repress transcriptional activation by HSF1 (Rabindran et al., 1994; Shi et al., 1998; Zou et al., 1998). Given the importance of CHIP’s Hsp70/Hsp90 interaction domain in the activation of HSF1, one explanation for CHIP-dependent activation of HSF1 is that CHIP disrupts Hsp/HSF1 complexes. To test this, we examined the effects of CHIP expression on these interactions. Either with or without crosslinkers, Hsp90/HSF1 interactions remained below the limits of our detection. However, in contrast to the expectation that CHIP would disrupt HSF1/Hsp70 complexes, we found that overexpression of CHIP led to stable Hsp70 interactions with the activated form of HSF1 (Figure 3B). These interactions were abolished by mutation of CHIP’s TPR domain, suggesting that this HSF1/Hsp70 complex required interactions between CHIP and Hsp70. In converse experiments, CHIP (as well as Hsp70) was also found in HSF1 immunoprecipitates (Figure 3C), and the stability of these complexes was similar to that of CHIP-dependent HSF1 trimerization (Figure 3A). We could also coimmunoprecipitate HSF1 with transiently transfected Myc-tagged CHIP (Figure 3D).

We have not identified direct interactions between CHIP and HSF1, and their association depends on the integrity of CHIP’s TPR domain, so these results are likely to indicate that CHIP, Hsp70 and activated HSF1 exist in an Hsp70-dependent ternary complex that resists HSF1 inactivation. To consider the mechanism of CHIP-dependent HSF1 activation in more detail, we evaluated the subcellular distribution of CHIP in the setting of cell stress. The activation of HSF1 in HeLa cells has been carefully considered, and requires cytoplasmic-to-nuclear translocation as an early event (Zuo et al., 1995; Cotto et al., 1997). A direct role for CHIP in HSF1 activation would require that a fraction of CHIP is nuclear-localizing. To examine this in more detail, we analyzed the subcellular distribution of CHIP and HSF1 after heat shock. In the unstimulated state, the majority of HSF1 is confined to the cytoplasm, as is the majority of CHIP (Figure 4A). As previously noted, HSF1 rapidly translocated to the nucleus after heat shock. Interestingly, there was an inversion of the cytoplasmic/nuclear ratio of CHIP as well, indicating that CHIP also relocalizes to the nucleus after heat shock. Both HSF1 and CHIP reaccumulate in the cytoplasm during recovery from heat shock. This observation was confirmed by immunofluorescence (Figure 4B), which demonstrates that cytoplasmic CHIP is largely localized in the region of the endoplasmic reticulum in unstressed conditions as previously described (Meacham et al., 2001). After heat shock, the cytoplasmic pool collapses and increased accumulation of CHIP is observed in punctate loci within the nucleus. During recovery, CHIP reaccumulates in the cytoplasm in a pattern similar to that observed in unstressed conditions.

The synchronous nucleocytoplasmic transport of HSF1 and CHIP led us to consider whether there is a direct association between endogenous CHIP and HSF1 in the setting of cellular stress. Indeed, consistent with temporal trends in localization, we could detect HSF1-containing CHIP immunocomplexes both before and after heat shock (Figure 4C). Notably, the HSF1–CHIP heterocomplex was detectable under conditions in which HSF1 is entirely nuclear localized. The kinetics of this association paralleled activation of HSF1 by heat shock, and during the recovery period HSF1 became largely dissociated from CHIP (Figure 4D). This suggested to us that CHIP may facilitate HSF1 translocation. To test this, we overexpressed CHIP and examined the subcellular localization of HSF1. Increasing levels of CHIP resulted in complete redistribution of HSF1 to the nucleus (Figure 4E), indicating that CHIP is indeed sufficient to induce HSF1 translocation. This observation provides a necessary mechanism for CHIP-dependent HSF1 activation.

Several arguments can be made to support the transcriptional competence of nuclear CHIP–HSF1 complexes: CHIP essentially induces quantitative HSF1 trimerization (Figure 3A), HSF1 DNA-binding activity is markedly increased in the presence of CHIP (Figure 1E) and CHIP-dependent upregulation of Hsp70 by HSF1 is potent and sustained (Figures 1B and 3C). To address this directly, we probed for the presence of CHIP in HSF1 complexes bound to the HSE. As previously observed (Figure 1E), overexpression of CHIP markedly enhanced HSF1 binding to an oligonucleotide containing its canonical sequence (Figure 4F). In addition to HSF1, we could detect both Hsp70 and CHIP in these complexes, indicating that the trimeric complex (Figures 3B and C) was competent for binding DNA. In addition, and consistent with the interaction of CHIP and activated HSF1 in the nucleus, endogenous CHIP was detected in DNA-bound HSF1-containing complexes that were transcriptionally activated by heat shock (Figure 4G). Interestingly, we detect less Hsp70 in DNA-bound HSF1 complexes after heat shock than when these complexes are directly induced by CHIP, presumably because Hsp70 is more likely to be sequestered by unfolded proteins under these circumstances and therefore is most likely to be dispensable for the DNA-binding activity of HSF1. The ability of CHIP to activate HSF1, and its interaction with HSF1 and participation in DNA-bound HSF1 complexes after heat shock, argues for a previously unanticipated role for CHIP in modulating the stress response that can best be tested by examining the cellular and physiologic consequences when CHIP is deficient.

Perinatal lethality in mice deficient in CHIP

If CHIP has such a central role in regulation of the general transcriptional stress response, as our studies suggest, then the cellular response to stress should be abated in the absence of CHIP. We generated mice deficient in CHIP to test this hypothesis. The first three exons of the CHIP gene were replaced with a PGK–Neo selection cassette by homologous recombination (Figure 5A). Germline transfer of the targeted allele was successful (Figure 5B) and deletion of CHIP expression was confirmed by western blotting (Figure 5C). Mice heterozygous at the CHIP locus were maintained on a mixed 129 Sv/Ev X C57BL/6 background. CHIP (–/–) mice were recovered in Mendelian frequencies from heterozygous crosses, indicating that CHIP is not required for essential developmental events in these mice (Figure 5D). However, 5% of mice were quickly cannibalized in the immediate peripartum period, most likely after perinatal death, and these mice could not be genotyped. We did not observe this rate of perinatal lethality in litters from wild-type crosses nor when wild-type males were bred with CHIP (+/–) females, excluding a maternal effect for the CHIP (+/–) genotype. We therefore presume the ungenotyped mice were predominantly CHIP (–/–). The remaining CHIP (–/–) offspring suckled normally; however, 20% of these mice also died in the peripartum period, which greatly exceeded the perinatal mortality of CHIP (+/+) or (+/–) mice (Figure 5E). This lethality was not rescued by transferring litters to wild-type foster mothers of mixed or FVB background. The dead CHIP (–/–) mice had no obvious morphologic abnormalities and had milk in their stomachs at necropsy. Normal surfactant production was confirmed by PAS staining in these mice.

The most striking structural lesion noted in the dead CHIP (–/–) mice was marked thymic atrophy (Figure 5F, upper panel). Histologic analysis demonstrated that thymic remnants were hypocellular and contained dense, pyknotic nuclear clumps, the hallmarks of massive thymocyte apoptosis (Figure 5F, lower panels). Surviving CHIP (–/–) mice had no obvious thymocyte abnormalities and were morphologically normal. Measurement of development of CD4(+) and CD8(+) by flow cytometry indicated that these cells developed normally in thymuses of CHIP (–/–) mice and that their numbers were not different from wild-type (data not shown), making a primary thymocyte abnormality unlikely. Peripartum thymocyte apoptosis is often a non-specific response to abnormal stress tolerance, and is due in part to increased levels of circulating corticosteroids (Tybulewicz et al., 1991). Based on additional studies described below, we interpret these results as indicators that CHIP (–/–) mice may adapt poorly to the stress of parturition.

CHIP is required for maximal Hsp70 induction and protection against stress-induced apoptosis

To address the role of CHIP in regulation of physiologic stress response pathways more directly, we isolated fibroblasts from CHIP (–/–) mice or wild-type littermates. CHIP (–/–) fibroblasts had typical morphologic, growth and viability characteristics under standard culture conditions (37°C). However, in comparison with wild-type fibroblasts, CHIP (–/–) fibroblasts exhibited markedly diminished responses to stress. For example, the ability of Hsp70 to be induced in response to heat shock was reduced by 50–60% in CHIP (–/–) fibroblasts (Figure 6A). This biochemical indication of impaired response to stress is similar to that which occurs in HSF1-null cells (McMillan et al., 1998), as would be expected from our overexpression studies (Figure 1). We also found that viability of CHIP (–/–) fibroblasts after heat shock (45°C) was sharply reduced compared with wild-type cells (Figure 6B). The stress-induced deficit in viability of cells lacking CHIP was accompanied by increased indices of apoptosis (Figure 6C), which is consistent with previous studies indicating that heat stress can activate an apoptotic cascade (Beere and Green, 2001). The stress-induced defect in CHIP (–/–) fibroblasts was not limited to a failure of tolerance to heat shock; cells lacking endogenous CHIP were also more susceptible to the amino acid analog l-canavanine, which incorporates into proteins during synthesis and induces misfolding (Figure 6D). We further examined whether CHIP (–/–) cells develop thermotolerance after preconditioning with a sublethal heat challenge (42°C), which is the primary defect in HSF1 (–/–) fibroblasts (McMillan et al., 1998). The loss of viability of CHIP (–/–) cells after thermal challenge was only partially rescued by preconditioning (Figure 6E). This suggested to us that CHIP may also be required for development of thermotolerance. The effect of CHIP on thermotolerance was more evident after prolonged (24 h) recovery after heat shock, at which time cell death after heat shock in the absence of preconditioning was maximal regardless of genotype (Figure 6F). To analyze biochemical markers of apoptosis, we measured caspase 3 activity after thermal stress with or without preconditioning. In comparison to wild-type cells, CHIP (–/–) cells markedly increased caspase 3 activity following thermal challenge at 45°C, and this increase is only modestly reduced by preconditioning (Figure 6G). Similarly, cleaved (and therefore activated) caspase 3 is only detected in CHIP (–/–) cells after heat shock, with or without preconditioning (Figure 6H). These in vitro studies support our interpretation of CHIP overexpression studies, and indicate that CHIP has an unanticipated central role in regulating the cellular response to stress.

Impaired responses to thermal challenge in CHIP (–/–) mice

We tested the physiologic significance of these in vitro studies in mice deficient in CHIP. Because we observed no obvious anatomic abnormalities in mature CHIP (–/–) mice, we considered the possibility that these mice might have a generally impaired response to physiologic stress, especially given the neonatal events that occurred in a subset of CHIP-deficient mice (Figure 5). Sensitivity to thermal challenge is a well-characterized physiologic stressor. Hsp70 levels were very low or undetectable in sham-treated mice, as is to be expected (Hotchkiss et al., 1993; Huang et al., 2001), regardless of their genotype. After sublethal elevation of core body temperatures, induction of Hsp70 was consistently lower in CHIP (–/–) mice than in wild-type mice (Figure 7A), which is similar to the responses of CHIP (–/–) cells in culture (Figure 6A). To test the physiologic response to hyperthermia, core body temperatures of mice were raised to 42°C for 15 min. All wild-type mice survived this challenge, and death during recovery was only 25%. (No deaths were noted after 24 h of recovery.) In marked contrast, 100% of CHIP (–/–) mice died rapidly, either during thermal challenge or immediately thereafter (Figure 7B). Necropsy revealed characteristic stress-responsive lesions. The primary abnormalities noted were edema, hyperemia and friability of the small intestine (Figure 7C), which are common lesions in humans suffering from heat stroke (Shibolet et al., 1967). In contrast to the normal villous architecture observed in heat-shocked wild-type mice (Figure 7D), marked blunting of villi was noted in CHIP (–/–) mice, as was thrombosis of subvillous arterioles (Figure 7E). Many cells within the damaged villi of CHIP (–/–) mice were shrunken with pyknotic nuclei (Figure 7F). TUNEL staining indicated that these cells were undergoing apoptosis (Figure 7G). It is most likely that the rapid death of CHIP (–/–) mice after thermal challenge is secondary to fluid shifts or endotoxemia from compromise of the gastrointestinal barrier. Apoptosis in heat-stressed CHIP (–/–) mice was not limited to the gastrointestinal tract. For example, we also observed sheets of dying splenocytes in CHIP-deficient mice after thermal challenge (compare Figures 7H and I). Most splenocytes of CHIP (–/–) mice were TUNEL-positive (Figure 7K) whereas this finding was rare in wild-type mice (Figure 7J).

Cell culture

COS7 cells were cultured as previously described (Connell et al., 2001). HSF1 (–/–) and HSF1 (+/+) murine embryonic fibroblasts were provided by Ivor Benjamin (McMillan et al., 1998; Xiao et al., 1999) and CHIP (–/–) and CHIP (+/+) fibroblasts were cultured according to standard protocols. Recombinant adenoviruses expressing CHIP or control viruses were constructed with the Ad-Easy system and cultures were routinely infected at a multiplicity of infection of five with an infection efficiency of >98%. Western blotting and immunoprecipitations were performed as previously described (Connell et al., 2001).

Reporter gene and electrophoretic mobility gel shift assays (EMSAs)

Hsp70 promoter:reporter constructs were provided by David Latchman (Stephanou et al., 1999). Reporter constructs were co-transfected in COS7 with a β-galactosidase reporter (to control for transfection efficiency), with or without plasmids expressing CHIP, CHIP mutants or other co-chaperones as indicated. For EMSA, a radiolabeled HSE oligonucleotide was incubated with nuclear extract from cells treated as indicated. The reaction mixture was incubated at room temperature for 20 min and fractionated on a 5% native polyacrylamide gel in 0.5× TBE buffer. To determine the specificity of the DNA–protein complexes, we performed competition assays using 50-fold molar excess of the unlabeled double-stranded HSE oligonucleotide (specific inhibitor) or excess of an unrelated NFκB oligonucleotide of comparable length (non-specific inhibitor). To characterize specific DNA-binding proteins, we incubated nuclear extracts with anti-HSF1 or anti-HSF2 antibody before adding probe.

Assays of apoptosis

CHIP (+/+) and CHIP (–/–) fibroblasts were incubated at 37°C, or heated at 42°C for 30 min, followed by 6 h recovery at 37°C and heated again at 45°C for 45 min (to stimulate maximal preconditioning), or only heated at 45°C for 45 min. Cells were incubated at 37°C as indicated in the figure legends before lysing. Viability was measured with the XTT Assay (Roche). Caspase 3 activity was measured using a fluorometric assay (Roche). Cleaved caspase 3, which is the activated form of the enzyme, was detected with cleaved caspase 3 (Asp175) antibody 9661 (Cell Signaling Technology), which detects only the active form of caspase 3.

Generation of the CHIP targeting construct and mutant mice

The CHIP targeting construct was generated using the pKO Scrambler vector (Stratagene), which contains a neomycin-resistance and thymidine kinase gene for positive and negative selection. Following electroporation into a 129 SvEv cell line, selected clones were analyzed for homologous recombination by Southern blotting. Three clones were injected into C57BL/6 blastocysts, and chimeric mice were bred to C57BL/6 females to achieve germline transmission of the targeted allele.

Thermal challenge

CHIP (–/–) mice and wild-type littermates of both sexes were maintained on a 129 SvEv X C57BL/6 background. Mice were anesthetized with 1.2% avertin (0 375 mg/kg body weight by intraperitoneal injection), immobilized and covered with a blanket, and placed underneath a 250 W infrared lamp. Core rectal temperatures were raised to 42°C over 15 min and maintained at this temperature transiently (for analysis of Hsp70 expression) or for 15 min (for mortality studies). After thermal challenge, mice were volume-repleted with intraperitoneal saline, and slowly returned to isothermal conditions on a warm blanket. Mice were directly observed for 24 h after thermal challenge. No deaths were observed in mice that survived beyond 24 h after thermal challenge.