Cell, Vol. 65, 363-366,
May 3, 1991,
Copyright
0 1991 by Cell Press
Heat Shock Factor and the Heat Shock Response Peter K. Sorger Department of Microbiology and Immunology University of California San Francisco, California 94143-0502
The induction of eukaryotic heat shock genes in response to a temperature upshift is mediated by the binding of a transcriptional activator, heat shock factor, to a short highly conserved DNA sequence known as the heat shock element. This review will discuss progress in several organisms in characterizing heat shock factor and will suggest preliminary answers to the following questions. How is the DNA-binding activity and transcriptional potency of heat shock factor regulated? How is autoregulation of heat shock protein (hsp) synthesis achieved? What is the mechanism by which heat shock is sensed? Heat Shock Elements and Their Interaction with Heat Shock Factor Heat shock elements are best described as contiguous arrays of variable numbers of the 5 bp sequence nGAAn arranged in alternating orientation (n denotes less strongly conserved nucleotides that nevertheless may be involved in important DNA-protein interactions; Figure 1; Xiao and Lis, 1988; Amin et al., 1988). At least two nGAAn units are needed for high affinity binding of heat shock factor in vitro, and these may be arranged either head-to-head (nGAAnnTTCn) or tail-to-tail (nTTCnnGAAn; Perisic et al., 1989). How can heat shock factor bind to these structurally distinct sites as well as to heat shock elements with larger numbers of 5 bp units? The answer may lie in the oligomerit nature of the heat shock factor protein (Figure 2). Heat shock factor from both Saccharomyces cerevisiae (SC-HSF) and Drosophila (D-HSF) associates to form protein trimers in solution and when bound to DNA (Perisic et al., 1989; Sorger and Nelson, 1989). It is not clear, however, whether heat shock factor exists in vivo primarily as a trimeric, hexameric, or possibly even larger complex. Each subunit of a D-HSF multimer is thought to bind to a single nGAAn unit, and the binding to successive units is highly cooperative (Xiao et al., 1991). The coiled-coil structure that has been proposed to form the interface between heat shock factor monomers is inherently threefold symmetric (see below). Thus, the binding of heat shock factor trimers to DNA (which is inherently two-fold symmetric) would require a flexible hinge between the trimerization and DNA-binding surfaces of individual monomers. Flexibility at this hinge may also be exploited in the binding of subunits to differently oriented nGAAn units. The binding of trimers to adjacent sites is also highly cooperative (Shuey and Parker, 1986): in vitro, the dissociation rate from DNA of a complex of two DNA-bound trimers is more than three orders of magnitude lower than that of a single trimer (Xiao et al., 1991). Thus, even when the molarity of sites exceeds that of the heat shock factor protein, heat shock factor preferentially forms large com-
Minireview
plexes, probably including hexamers (Sorger and Nelson, 1989; Clos et al., 1990). The formation of these large complexes is almost certainly important for the activation of heat shock genes in vivo. Although a heat shock element containing three nGAAn units binds D-HSF trimers tightly in vitro, arrays of more than three units must be introduced into the Drosophila genome to stimulate high-level transcription (Xiao and Lis, 1988). In mammalian cells, only heat shock elements containing extended nGAAn arrays activate transcription when located far from the TATA box (Benz and Pelham, 1986). These results imply that transcriptionally active complexes contain six or more heat shock factor monomers. Architecture of the Heat Shock Factor Protein The gene encoding heat shock factor was first isolated from S. cerevisiae. It is essential for viability at all temperatures, and SC-HSF has no close relatives with similar functions (Wiederrecht et al., 1988; Sorger and Pelham, 1988). Heat shock factor genes have also been isolated from Drosophila (Clos et al., 1990), tomato cells (Scharf et al., 1990) and the yeast Kluyveromyces lactis (KI-HSF; Jakobsen and Pelham, 1991). Tomato cells contain two or more proteins with the properties of heat shock factor, but the
caTCc GAA ccTCt GAA *IIballIballi$nali ..-. -.--).---._ --)a
tTTCt GA cTTCc .-_.__*-
--I
5 bp Figure
1. Heat Shock
Elements
Are Arrays
of 5 bp Units
The sequence shown is a heat shock element from the Drosophila hsp83 promoter. Arrows indicate the orientation of the units, and dashed lines denote imperfect matches to the consensus nGAAn sequence.
Figure 2. Model for the Interaction Different Arrays of 5 bp Units
of Heat Shock
Factor Trimers
with
The foot represents the DNA-binding domain of heat shock factor, and the legs join the subunits of the trimer. The figure was kindly provided by Dr. John Lis and is reprinted with permission of Cold Spring Harbor Laboratory Press from Morimoto et al. (1990).
Cell 364
significance of this for the regulation of the heat shock response is not yet known. Perhaps the various activities that appear to be mediated by heat shock factor in budding yeasts are divided among several polypeptides in higher eukaryotes. Despite the strong phylogenetic conservation of the heat shock element sequence, heat shock factor proteins from different species have only limited sequence similarity. For example, SC-HSF and KI-HSF share only 18% amino acid identity. The similarity is confined to the DNA binding domain and to the region that is involved in trimerization. Nevertheless, KI-HSF can substitute for SC-HSF in S. cerevisiae, and fusions that link the N-terminus of one factor with the C-terminus of the other are correctly regulated (Jakobsen and Pelham, 1991). This suggests that, despite the divergence in primary structure, the domain organization and three-dimensional folds of the two heat shock factors are similar. The heat shock factor DNA-binding domain contains a short but good match to the putative DNA recognition helix of bacterial sigm,a factors (Clos et al., 1990), but has no close relatives among known eukaryotic transcriptional activators (Wiederrecht et al., 1988). The discovery of a highly homologous domain in a yeast protein not thought to play a role in the heat shock response (SLFl , a flocculation suppressor; Fujita et al., 1989) suggests, however, that the heat shock factor DNA-binding domain may be a member of a family of domains with similar sequences. The region of heat shock factor involved in trimerization lies C-terminal to the DNA-binding domain and has been proposed to form a three-stranded a-helical coiled coil (Sorger and Nelson, 1989). Although the residues postulated to form the interface between helices are only partially conserved among different organisms, the distributions of hydrophobic and charged residues are nearly identical, implyingthatverysimilarstructuresareinvolved. Activation of Heat Shock Factor by Heat Shock Upon heat shock, a preexisting pool of unactivated heat
AUTOREGULATION IN ANIMAL CELLS:
0
shock factor is converted into a form capable of efficiently stimulating transcription. In Drosophila, human, and S. pombe cells, heat shock factor binds to DNA only after heat shock (see Gallo et al., 1991, and references therein). This suggests a simple mechanism for induction in which acquisition of the ability to bind DNA allows heat shock factor to interact with and activate heat shock promoters. In S. cerevisiae, however, SC-HSF is bound to DNA both before and after heat shock. Both hu-HSF and SC-HSF become highly phosphorylated following heat shock, and in S. cerevisiae the transcriptional activity of the factor is closely correlated over a range of temperatures with the extent of its phosphorylation. These observations suggest that, in animal cells, the activation of heat shock factor involves two mechanistically (although not necessarily temporally) distinct steps: first, the induction of DNA binding to create a heat shock factor complex bound to heat shock promoters, and second, the phosphorylation of heat shock factor to create a complex with high transcriptional activity (Figure 3; Larson et al., 1988). In S. cerevisiae, induction by heat shock would involve only the second step. Heat shock factor prepared from unshocked animal cells can be induced to bind DNA in vitro by exposing cell extracts to elevated temperatures or to reagents that favor the dissociation and denaturation of protein complexes (Larson et al., 1988). However, recombinant D-HSF produced in E. coli (but not in Xenopus oocytes) binds to heat shock elements with high affinity in the absence of treatment with heat or denaturants (Clos et al., 1990). This suggests that heat shock factor may interact with one or more negative regulators found in eukaryotic cells and that the interactions can be disrupted in vitro by conditions that mimic the effects of heat shock (Clos et al., 1990). If we speculate that these putative negative regulators are hsps, the following model for autoregulation of the heat shock response can be formulated. Under normal growth conditions, hsps bind to and repress heat shock factor
AUTOREGULATION IN S. CEREVISIAE:
Figure 3. Speculative Model for the Activation of Heat Shock Factor by Heat Shock in Human and Drosophila Cells and in the Yeast S. cerevisiae Despite its conceptual simplicity, readers should note that this model has not been fully confirmed experimentally. See text for details.
Minireview 365
activity. These hsps may include hsp70, which has been implicated in the homeostasis of the Drosophila heat shock response (DiDomenico et al., 1982) and hsp90, which interacts with steroid hormone receptors and regulates their transcriptional activity (e.g., Picard et al., 1990). During heat shock, competition with high levels of thermally damaged proteins for binding to hsps causes the dissociation of heat shock factor-hsp complexes and a consequent increase in the DNA-binding affinity of the factor. DNA-bound heat shock factor then directs increased hsp synthesis until levels are sufficiently high to result in the reassociation of hsps and heat shock factor and the reestablishment of the repressed state (similar models are discussed in Morimoto et al., 1990). How might the DNA-binding affinity of heat shock factor be regulated? Monomeric forms of SC-HSF and D-HSF (generated in vitro bytruncation of the protein) bind to DNA much more poorly than do multimers (Sorger and Nelson, 1989; Closet al., 1990). Thus, the induction of heat shock factor DNA binding in animal cells may involve a transition from a monomeric (unactivated) to an oligomeric state. Although this idea remains speculative, it is consistent with the observation that antibodies directed against D-HSF, which would be expected to link monomers, induce DNA binding in extracts from unshocked Drosophila cells (Zimarino et al., 1990). Hsps could regulate the oligomerization state of heat shock factor (and hence its ability to bind to DNA) by masking the trimerization surface. More complex roles for hsps can also be envisioned. By analogy to the cooperative roles played by the groEL and groES hsps in protein assembly (e.g., Goloubinoff et al., 1989) several different hsps might interact with heat shock factor. These hsps might regulate heat shock factor negatively in normal conditions, facilitate the assembly of heat shock factor into oligomeric complexes during induction, return heat shock factor to an inactive configuration during recovery, or regulate the kinases and phosphatases that act on heat shock factor. It must be emphasized, however, that, despite the conceptual simplicity of the model presented in Figure3 there is no data directly demonstrating that hsps regulate the activity of heat shock factor. The results discussed above are also consistent with the possibility that conformation of heat shock factor is affected by heat shock, and that temperature-dependent conformational changes control the ability of the factor to bind DNA. Because heat shock factor is bound to DNA under all conditions in S. cerevisiae, it is not clear whether changes in the oligomerization state of SC-HSF play a role in regulation. Genetic experiments in S. cerevisiae and K. lactis have demonstrated that the transcriptional activity of heat shock factor in budding yeasts is negatively regulated. These heat shock factors can be strongly activated by the deletion of several different regions of the heat shock factor protein (Jakobsen and Pelham, 1991, and references therein). The integrity of the trimerization region, and possibly also part of the DNA-binding domain, is essential for the maintenance of a repressed state in the absence of heat shock. The mechanism by which deletions in heat shock factor enhance its transcriptional activity is un-
known; it could involve the disruption eitherof intramolecular interactions or of intermolecular interactions with negative regulators. The data suggest, however, that heat shock may act to unmask a normally repressed transcriptional activation domain, and that the masking of this activator in the absence of heat shock involves the highly conserved structural core of the heat shock factor protein, which includes the trimerization domain (Jakobsen and Pelham, 1991). In S. cerevisiae, the only demonstrable biochemical difference between heat shock factor from shocked and unshocked cells is the extent of its phosphorylation (Sorger and Pelham, 1988), and a correlation between transcriptional activation and heat shock-dependent phosphorylation of heat shock factor has been observed in all organisms examined. For example, hsp70 transcription is not induced by heat shock in murine erythroid leukemia cells, and, although heat shock factor binds to DNAafter a temperature upshift, it does not become phosphorylated in these cells (Hensold et al., 1990). Thus, the binding of murine heat shock factor to DNA appears to be insufficient for transcriptional activation. Although these observations are only correlative, one can speculate that phosphorylation is directly involved in increasing the ability of heat shock factor to enhance transcription and constitutes a second step in the pathway of heat shock factor activation (Figure 3). To test this idea, it will be necessary to identify and mutate the residues in heat shock factor that become modified after heat shock. Transcriptional Activation by Heat Shock Factor The classical heat shock response is inherently transient, characterized by a rapid 30- to lOO-fold increase in heat shock factor activity that persists for less than 1 hr. If heat shock factor mediates only this transient response, why is HSFan essential gene in S. cerevisiae at all temperatures? In budding yeast, heat shock genes must be expressed continuously for normal growth (Craig and Jacobsen, 1984), and although a variety of transcription factors regulate their expression, several lines of evidence suggest that heat shock factor plays an important role. First, the deletion of heat shock elements significantly decreases the transcriptional activity of a yeast hsp70 (Park and Craig, 1989) and an hsp82 (McDaniel et al., 1989) promoter in unshocked cells, and synthetic heat shock elements function as promoters at low temperatures. Second, the overexpression of heat shock factor in wild-type cells in the absence of heat shock increases the level of a major species of hsp70 (Sorger and Pelham, 1988). Third, in addition to the classical transient activation, the exposure of budding yeast cells to elevated temperatures also results in sustained changes in heat shock factor activity. Whereas the transient changes are regulated over a narrow range of temperatures above 34’%, sustained changes in activity are regulated over the full range of temperatures at which S. cerevisiae can grow. The transient and sustained activities of heat shock factor appear to be mediated by physically separable regions of the heat shock factor polypeptide and may be regulated independently in response to different stimuli (Sorger, 1990).
Cell 366
Thus, a continuous requirement for heat shock factor may reflect its involvement in the regulation of hsp synthesis in the absence of heat shock. One intriguing possibility is that the factor can be activated not only by stressdamaged proteins, but also by hsp substrates produced in normally growing cells. It remains to be determined whether heat shock factor has a function in unstressed animal cells, but there are indications that heat shock factor can be induced to bind to DNA by conditions that are not conventionally stressful (Mezger et al., 1989; Theodorakis et al., 1989). Unresolved Issues The mechanism by which heat shock factor activates transcription is unknown. In unshocked Drosophila cells, a molecule of RNA polymerase II is present near the start of the hsp70 promoter and has transcribed about 25 nucleotides (Rougvie and Lis, 1988). This suggests that the role of D-HSF may be to overcome a block imposed on elongation by a committed polymerase rather than to increase the rate of initiation. Several features in Figure 3 await experimental confirmation. In particular, it is not yet known whether the DNAbinding activity of heat shock factor is regulated by oligomerization nor whether hsps bind to heat shock factor. A direct demonstration of a role for phosphorylation in enhancing the transcriptional activity of heat shock factor is also lacking. At a more fundamental level, the most important unresolved problem is determination of the mechanisms by which heat shock is sensed and autoregulation achieved. If the current model is correct in outline if not in detail, then we can imagine two possibilities. The sensor may consist of heat shock factor and the negative regulators to which it binds; if these negative regulators are hsps, then a simple model for autoregulation of the heat shock response can be formulated, as described above. In this case, the kinases and phosphatases that modify heat shock factor would not be regulated by heat shock, and increased phosphorylation of the factor would reflect conformational changes that increase its accessibility as a kinase substrate (Jakobsen and Pelham, 1991). If increased phosphorylation is a consequence rather than the cause of heat shock factor activation, phosphorylation might nevertheless be involved in regulating the half-life of the activated state or even in down-regulating activity during recovery. The second possibility is that the kinases and phosphatases that modify heat shock factor constitute critical components of the heat shock sensor and transduce an as yet unidentified stress signal. In this view, the kinase would be activated (or phosphatase inactivated) by heat shock, causing increased phosphorylation of heat shock factor and the consequent stimulation of its ability to activate transcription. A major challenge in the future will be to determine which of these views is correct, or more likely, to assess their relative importance in a multistep pathway of heat shock factor activation.
Amin, J., Ananthan, 3769. Bienz,
J., and Voellmy,
M., and Pelham,
Ft. (1966).
H. R. B. (1966).
Mol. Cell. Biol. 8,3761-
Cell 45, 753-760
Clos, J., Westwood, J. T., Becker, P. B., Wilson, Wu, C. (1990). Cell 63, 1065-1097. Craig,
E. A., and Jacobsen,
DiDomenico, Natl. Acad.
K. (1984).
Gallo, G. J., Shuetz, 11, 261-266.
Hensold, Kingston,
T. J., and Kingston,
P., Gatenby,
S. L. (1962).
Y., Matsumoto,
Ft. E. (1991).
A. A., and Lorimer,
B., and Pelham,
H. R. 8. (1991).
Larson, J. S., Shuetz, T. J., and Kingston, 372-375. Erratum 336, 164.
Proc.
S., and
Mol. Cell. Biol.
G. H. (1969). Nature
J. O., Hunt, C. Ft., Calderwood, S. K., Housman, Ft. E. (1990). Mol. Cell. Biol. 70, 1600-1606.
Jakobsen,
K., and
Cell 38, 641-849.
8. J., Bugaisky, G. E., and Lindquist, Sci. USA 79, 6161-6165.
Fujita, A., Kikuchi, Y., Kuhara, S., Misumi, Kobayashi, H. (1969). Gene 85321-326.
Goloubinoff, 44-47.
S., Lambert,
EMBO
337,
D. E., and
J. 70, 369-375.
R. E. (1966).
Nature
335,
McDaniel, D., Caplan, A. J., Lee, M.-S., Adams, C. C., Fishel, B. R., Gross, D. S., and Garrard, W. T. (1969). Mol. Cell. Biol. 9,4789-4798. Mezger, V., Bensaude, 3888-3896.
O., and Morange,
M. (1989).
Mol. Cell. Biol. 9,
Morimoto, R. I., Tissieres, A., and Georgopoulos, C., eds. (1990). Stress Proteins in Biology and Medicine (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Park,
H., and Craig,
Perisic,
E. A. (1989).
Mol. Cell. Biol. 9, 2025-2033.
O., Xiao, H., and Lis, J. T. (1989).
Cell 59, 797-806.
Picard, D., Khursheed, B., Garabedian, M. J., Fortin, M. G., Lindquist, S., and Yamamoto, K. R. (1990). Nature 348, 166-168. Rougvie,
A. E., and Lis, J. T. (1988).
Scharf, EMBO
K.-D., Rose, S., Zott, J. 9, 4495-4501.
Shuey,
D. J., and Parker,
Sorger,
P. K. (1990).
Cell 54, 795-804.
W., Schoff,
F., and Nover,
L. (1990).
C. S. (1986). J. Biol. Chem. 261,7934-7940.
Cell 62, 793-805.
Sorger,
P. K., and Nelson,
H. C. M. (1989).
Cell 59, 807-813.
Sorger,
P. K., and Pelham,
H. R. B. (1988).
Cell 54, 855-864.
Theodorakis, N. G., Zand, D. J., Kotzbauer, P. T., Williams, Morimoto, R. I. (1989). Mol. Cell. Biol. 9, 3166-3173. Wiederrecht,
G., Seto, D., and Parker,
Xiao, H., and Lis, J. T. (1988). Xiao, H., Perisic, Zimarino,
Science
C. S. (1968).
Cell 54,841-853.
239, 1139-1142.
O., and Lis, J. T. (1991).
V., Wilson,
G. T., and
S., and Wu, C. (1990).
Cell 64, 585-593. Science
249, 546-549.
May 3, 1991,
Copyright
0 1991 by Cell Press
Heat Shock Factor and the Heat Shock Response Peter K. Sorger Department of Microbiology and Immunology University of California San Francisco, California 94143-0502
The induction of eukaryotic heat shock genes in response to a temperature upshift is mediated by the binding of a transcriptional activator, heat shock factor, to a short highly conserved DNA sequence known as the heat shock element. This review will discuss progress in several organisms in characterizing heat shock factor and will suggest preliminary answers to the following questions. How is the DNA-binding activity and transcriptional potency of heat shock factor regulated? How is autoregulation of heat shock protein (hsp) synthesis achieved? What is the mechanism by which heat shock is sensed? Heat Shock Elements and Their Interaction with Heat Shock Factor Heat shock elements are best described as contiguous arrays of variable numbers of the 5 bp sequence nGAAn arranged in alternating orientation (n denotes less strongly conserved nucleotides that nevertheless may be involved in important DNA-protein interactions; Figure 1; Xiao and Lis, 1988; Amin et al., 1988). At least two nGAAn units are needed for high affinity binding of heat shock factor in vitro, and these may be arranged either head-to-head (nGAAnnTTCn) or tail-to-tail (nTTCnnGAAn; Perisic et al., 1989). How can heat shock factor bind to these structurally distinct sites as well as to heat shock elements with larger numbers of 5 bp units? The answer may lie in the oligomerit nature of the heat shock factor protein (Figure 2). Heat shock factor from both Saccharomyces cerevisiae (SC-HSF) and Drosophila (D-HSF) associates to form protein trimers in solution and when bound to DNA (Perisic et al., 1989; Sorger and Nelson, 1989). It is not clear, however, whether heat shock factor exists in vivo primarily as a trimeric, hexameric, or possibly even larger complex. Each subunit of a D-HSF multimer is thought to bind to a single nGAAn unit, and the binding to successive units is highly cooperative (Xiao et al., 1991). The coiled-coil structure that has been proposed to form the interface between heat shock factor monomers is inherently threefold symmetric (see below). Thus, the binding of heat shock factor trimers to DNA (which is inherently two-fold symmetric) would require a flexible hinge between the trimerization and DNA-binding surfaces of individual monomers. Flexibility at this hinge may also be exploited in the binding of subunits to differently oriented nGAAn units. The binding of trimers to adjacent sites is also highly cooperative (Shuey and Parker, 1986): in vitro, the dissociation rate from DNA of a complex of two DNA-bound trimers is more than three orders of magnitude lower than that of a single trimer (Xiao et al., 1991). Thus, even when the molarity of sites exceeds that of the heat shock factor protein, heat shock factor preferentially forms large com-
Minireview
plexes, probably including hexamers (Sorger and Nelson, 1989; Clos et al., 1990). The formation of these large complexes is almost certainly important for the activation of heat shock genes in vivo. Although a heat shock element containing three nGAAn units binds D-HSF trimers tightly in vitro, arrays of more than three units must be introduced into the Drosophila genome to stimulate high-level transcription (Xiao and Lis, 1988). In mammalian cells, only heat shock elements containing extended nGAAn arrays activate transcription when located far from the TATA box (Benz and Pelham, 1986). These results imply that transcriptionally active complexes contain six or more heat shock factor monomers. Architecture of the Heat Shock Factor Protein The gene encoding heat shock factor was first isolated from S. cerevisiae. It is essential for viability at all temperatures, and SC-HSF has no close relatives with similar functions (Wiederrecht et al., 1988; Sorger and Pelham, 1988). Heat shock factor genes have also been isolated from Drosophila (Clos et al., 1990), tomato cells (Scharf et al., 1990) and the yeast Kluyveromyces lactis (KI-HSF; Jakobsen and Pelham, 1991). Tomato cells contain two or more proteins with the properties of heat shock factor, but the
caTCc GAA ccTCt GAA *IIballIballi$nali ..-. -.--).---._ --)a
tTTCt GA cTTCc .-_.__*-
--I
5 bp Figure
1. Heat Shock
Elements
Are Arrays
of 5 bp Units
The sequence shown is a heat shock element from the Drosophila hsp83 promoter. Arrows indicate the orientation of the units, and dashed lines denote imperfect matches to the consensus nGAAn sequence.
Figure 2. Model for the Interaction Different Arrays of 5 bp Units
of Heat Shock
Factor Trimers
with
The foot represents the DNA-binding domain of heat shock factor, and the legs join the subunits of the trimer. The figure was kindly provided by Dr. John Lis and is reprinted with permission of Cold Spring Harbor Laboratory Press from Morimoto et al. (1990).
Cell 364
significance of this for the regulation of the heat shock response is not yet known. Perhaps the various activities that appear to be mediated by heat shock factor in budding yeasts are divided among several polypeptides in higher eukaryotes. Despite the strong phylogenetic conservation of the heat shock element sequence, heat shock factor proteins from different species have only limited sequence similarity. For example, SC-HSF and KI-HSF share only 18% amino acid identity. The similarity is confined to the DNA binding domain and to the region that is involved in trimerization. Nevertheless, KI-HSF can substitute for SC-HSF in S. cerevisiae, and fusions that link the N-terminus of one factor with the C-terminus of the other are correctly regulated (Jakobsen and Pelham, 1991). This suggests that, despite the divergence in primary structure, the domain organization and three-dimensional folds of the two heat shock factors are similar. The heat shock factor DNA-binding domain contains a short but good match to the putative DNA recognition helix of bacterial sigm,a factors (Clos et al., 1990), but has no close relatives among known eukaryotic transcriptional activators (Wiederrecht et al., 1988). The discovery of a highly homologous domain in a yeast protein not thought to play a role in the heat shock response (SLFl , a flocculation suppressor; Fujita et al., 1989) suggests, however, that the heat shock factor DNA-binding domain may be a member of a family of domains with similar sequences. The region of heat shock factor involved in trimerization lies C-terminal to the DNA-binding domain and has been proposed to form a three-stranded a-helical coiled coil (Sorger and Nelson, 1989). Although the residues postulated to form the interface between helices are only partially conserved among different organisms, the distributions of hydrophobic and charged residues are nearly identical, implyingthatverysimilarstructuresareinvolved. Activation of Heat Shock Factor by Heat Shock Upon heat shock, a preexisting pool of unactivated heat
AUTOREGULATION IN ANIMAL CELLS:
0
shock factor is converted into a form capable of efficiently stimulating transcription. In Drosophila, human, and S. pombe cells, heat shock factor binds to DNA only after heat shock (see Gallo et al., 1991, and references therein). This suggests a simple mechanism for induction in which acquisition of the ability to bind DNA allows heat shock factor to interact with and activate heat shock promoters. In S. cerevisiae, however, SC-HSF is bound to DNA both before and after heat shock. Both hu-HSF and SC-HSF become highly phosphorylated following heat shock, and in S. cerevisiae the transcriptional activity of the factor is closely correlated over a range of temperatures with the extent of its phosphorylation. These observations suggest that, in animal cells, the activation of heat shock factor involves two mechanistically (although not necessarily temporally) distinct steps: first, the induction of DNA binding to create a heat shock factor complex bound to heat shock promoters, and second, the phosphorylation of heat shock factor to create a complex with high transcriptional activity (Figure 3; Larson et al., 1988). In S. cerevisiae, induction by heat shock would involve only the second step. Heat shock factor prepared from unshocked animal cells can be induced to bind DNA in vitro by exposing cell extracts to elevated temperatures or to reagents that favor the dissociation and denaturation of protein complexes (Larson et al., 1988). However, recombinant D-HSF produced in E. coli (but not in Xenopus oocytes) binds to heat shock elements with high affinity in the absence of treatment with heat or denaturants (Clos et al., 1990). This suggests that heat shock factor may interact with one or more negative regulators found in eukaryotic cells and that the interactions can be disrupted in vitro by conditions that mimic the effects of heat shock (Clos et al., 1990). If we speculate that these putative negative regulators are hsps, the following model for autoregulation of the heat shock response can be formulated. Under normal growth conditions, hsps bind to and repress heat shock factor
AUTOREGULATION IN S. CEREVISIAE:
Figure 3. Speculative Model for the Activation of Heat Shock Factor by Heat Shock in Human and Drosophila Cells and in the Yeast S. cerevisiae Despite its conceptual simplicity, readers should note that this model has not been fully confirmed experimentally. See text for details.
Minireview 365
activity. These hsps may include hsp70, which has been implicated in the homeostasis of the Drosophila heat shock response (DiDomenico et al., 1982) and hsp90, which interacts with steroid hormone receptors and regulates their transcriptional activity (e.g., Picard et al., 1990). During heat shock, competition with high levels of thermally damaged proteins for binding to hsps causes the dissociation of heat shock factor-hsp complexes and a consequent increase in the DNA-binding affinity of the factor. DNA-bound heat shock factor then directs increased hsp synthesis until levels are sufficiently high to result in the reassociation of hsps and heat shock factor and the reestablishment of the repressed state (similar models are discussed in Morimoto et al., 1990). How might the DNA-binding affinity of heat shock factor be regulated? Monomeric forms of SC-HSF and D-HSF (generated in vitro bytruncation of the protein) bind to DNA much more poorly than do multimers (Sorger and Nelson, 1989; Closet al., 1990). Thus, the induction of heat shock factor DNA binding in animal cells may involve a transition from a monomeric (unactivated) to an oligomeric state. Although this idea remains speculative, it is consistent with the observation that antibodies directed against D-HSF, which would be expected to link monomers, induce DNA binding in extracts from unshocked Drosophila cells (Zimarino et al., 1990). Hsps could regulate the oligomerization state of heat shock factor (and hence its ability to bind to DNA) by masking the trimerization surface. More complex roles for hsps can also be envisioned. By analogy to the cooperative roles played by the groEL and groES hsps in protein assembly (e.g., Goloubinoff et al., 1989) several different hsps might interact with heat shock factor. These hsps might regulate heat shock factor negatively in normal conditions, facilitate the assembly of heat shock factor into oligomeric complexes during induction, return heat shock factor to an inactive configuration during recovery, or regulate the kinases and phosphatases that act on heat shock factor. It must be emphasized, however, that, despite the conceptual simplicity of the model presented in Figure3 there is no data directly demonstrating that hsps regulate the activity of heat shock factor. The results discussed above are also consistent with the possibility that conformation of heat shock factor is affected by heat shock, and that temperature-dependent conformational changes control the ability of the factor to bind DNA. Because heat shock factor is bound to DNA under all conditions in S. cerevisiae, it is not clear whether changes in the oligomerization state of SC-HSF play a role in regulation. Genetic experiments in S. cerevisiae and K. lactis have demonstrated that the transcriptional activity of heat shock factor in budding yeasts is negatively regulated. These heat shock factors can be strongly activated by the deletion of several different regions of the heat shock factor protein (Jakobsen and Pelham, 1991, and references therein). The integrity of the trimerization region, and possibly also part of the DNA-binding domain, is essential for the maintenance of a repressed state in the absence of heat shock. The mechanism by which deletions in heat shock factor enhance its transcriptional activity is un-
known; it could involve the disruption eitherof intramolecular interactions or of intermolecular interactions with negative regulators. The data suggest, however, that heat shock may act to unmask a normally repressed transcriptional activation domain, and that the masking of this activator in the absence of heat shock involves the highly conserved structural core of the heat shock factor protein, which includes the trimerization domain (Jakobsen and Pelham, 1991). In S. cerevisiae, the only demonstrable biochemical difference between heat shock factor from shocked and unshocked cells is the extent of its phosphorylation (Sorger and Pelham, 1988), and a correlation between transcriptional activation and heat shock-dependent phosphorylation of heat shock factor has been observed in all organisms examined. For example, hsp70 transcription is not induced by heat shock in murine erythroid leukemia cells, and, although heat shock factor binds to DNAafter a temperature upshift, it does not become phosphorylated in these cells (Hensold et al., 1990). Thus, the binding of murine heat shock factor to DNA appears to be insufficient for transcriptional activation. Although these observations are only correlative, one can speculate that phosphorylation is directly involved in increasing the ability of heat shock factor to enhance transcription and constitutes a second step in the pathway of heat shock factor activation (Figure 3). To test this idea, it will be necessary to identify and mutate the residues in heat shock factor that become modified after heat shock. Transcriptional Activation by Heat Shock Factor The classical heat shock response is inherently transient, characterized by a rapid 30- to lOO-fold increase in heat shock factor activity that persists for less than 1 hr. If heat shock factor mediates only this transient response, why is HSFan essential gene in S. cerevisiae at all temperatures? In budding yeast, heat shock genes must be expressed continuously for normal growth (Craig and Jacobsen, 1984), and although a variety of transcription factors regulate their expression, several lines of evidence suggest that heat shock factor plays an important role. First, the deletion of heat shock elements significantly decreases the transcriptional activity of a yeast hsp70 (Park and Craig, 1989) and an hsp82 (McDaniel et al., 1989) promoter in unshocked cells, and synthetic heat shock elements function as promoters at low temperatures. Second, the overexpression of heat shock factor in wild-type cells in the absence of heat shock increases the level of a major species of hsp70 (Sorger and Pelham, 1988). Third, in addition to the classical transient activation, the exposure of budding yeast cells to elevated temperatures also results in sustained changes in heat shock factor activity. Whereas the transient changes are regulated over a narrow range of temperatures above 34’%, sustained changes in activity are regulated over the full range of temperatures at which S. cerevisiae can grow. The transient and sustained activities of heat shock factor appear to be mediated by physically separable regions of the heat shock factor polypeptide and may be regulated independently in response to different stimuli (Sorger, 1990).
Cell 366
Thus, a continuous requirement for heat shock factor may reflect its involvement in the regulation of hsp synthesis in the absence of heat shock. One intriguing possibility is that the factor can be activated not only by stressdamaged proteins, but also by hsp substrates produced in normally growing cells. It remains to be determined whether heat shock factor has a function in unstressed animal cells, but there are indications that heat shock factor can be induced to bind to DNA by conditions that are not conventionally stressful (Mezger et al., 1989; Theodorakis et al., 1989). Unresolved Issues The mechanism by which heat shock factor activates transcription is unknown. In unshocked Drosophila cells, a molecule of RNA polymerase II is present near the start of the hsp70 promoter and has transcribed about 25 nucleotides (Rougvie and Lis, 1988). This suggests that the role of D-HSF may be to overcome a block imposed on elongation by a committed polymerase rather than to increase the rate of initiation. Several features in Figure 3 await experimental confirmation. In particular, it is not yet known whether the DNAbinding activity of heat shock factor is regulated by oligomerization nor whether hsps bind to heat shock factor. A direct demonstration of a role for phosphorylation in enhancing the transcriptional activity of heat shock factor is also lacking. At a more fundamental level, the most important unresolved problem is determination of the mechanisms by which heat shock is sensed and autoregulation achieved. If the current model is correct in outline if not in detail, then we can imagine two possibilities. The sensor may consist of heat shock factor and the negative regulators to which it binds; if these negative regulators are hsps, then a simple model for autoregulation of the heat shock response can be formulated, as described above. In this case, the kinases and phosphatases that modify heat shock factor would not be regulated by heat shock, and increased phosphorylation of the factor would reflect conformational changes that increase its accessibility as a kinase substrate (Jakobsen and Pelham, 1991). If increased phosphorylation is a consequence rather than the cause of heat shock factor activation, phosphorylation might nevertheless be involved in regulating the half-life of the activated state or even in down-regulating activity during recovery. The second possibility is that the kinases and phosphatases that modify heat shock factor constitute critical components of the heat shock sensor and transduce an as yet unidentified stress signal. In this view, the kinase would be activated (or phosphatase inactivated) by heat shock, causing increased phosphorylation of heat shock factor and the consequent stimulation of its ability to activate transcription. A major challenge in the future will be to determine which of these views is correct, or more likely, to assess their relative importance in a multistep pathway of heat shock factor activation.
Amin, J., Ananthan, 3769. Bienz,
J., and Voellmy,
M., and Pelham,
Ft. (1966).
H. R. B. (1966).
Mol. Cell. Biol. 8,3761-
Cell 45, 753-760
Clos, J., Westwood, J. T., Becker, P. B., Wilson, Wu, C. (1990). Cell 63, 1065-1097. Craig,
E. A., and Jacobsen,
DiDomenico, Natl. Acad.
K. (1984).
Gallo, G. J., Shuetz, 11, 261-266.
Hensold, Kingston,
T. J., and Kingston,
P., Gatenby,
S. L. (1962).
Y., Matsumoto,
Ft. E. (1991).
A. A., and Lorimer,
B., and Pelham,
H. R. 8. (1991).
Larson, J. S., Shuetz, T. J., and Kingston, 372-375. Erratum 336, 164.
Proc.
S., and
Mol. Cell. Biol.
G. H. (1969). Nature
J. O., Hunt, C. Ft., Calderwood, S. K., Housman, Ft. E. (1990). Mol. Cell. Biol. 70, 1600-1606.
Jakobsen,
K., and
Cell 38, 641-849.
8. J., Bugaisky, G. E., and Lindquist, Sci. USA 79, 6161-6165.
Fujita, A., Kikuchi, Y., Kuhara, S., Misumi, Kobayashi, H. (1969). Gene 85321-326.
Goloubinoff, 44-47.
S., Lambert,
EMBO
337,
D. E., and
J. 70, 369-375.
R. E. (1966).
Nature
335,
McDaniel, D., Caplan, A. J., Lee, M.-S., Adams, C. C., Fishel, B. R., Gross, D. S., and Garrard, W. T. (1969). Mol. Cell. Biol. 9,4789-4798. Mezger, V., Bensaude, 3888-3896.
O., and Morange,
M. (1989).
Mol. Cell. Biol. 9,
Morimoto, R. I., Tissieres, A., and Georgopoulos, C., eds. (1990). Stress Proteins in Biology and Medicine (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Park,
H., and Craig,
Perisic,
E. A. (1989).
Mol. Cell. Biol. 9, 2025-2033.
O., Xiao, H., and Lis, J. T. (1989).
Cell 59, 797-806.
Picard, D., Khursheed, B., Garabedian, M. J., Fortin, M. G., Lindquist, S., and Yamamoto, K. R. (1990). Nature 348, 166-168. Rougvie,
A. E., and Lis, J. T. (1988).
Scharf, EMBO
K.-D., Rose, S., Zott, J. 9, 4495-4501.
Shuey,
D. J., and Parker,
Sorger,
P. K. (1990).
Cell 54, 795-804.
W., Schoff,
F., and Nover,
L. (1990).
C. S. (1986). J. Biol. Chem. 261,7934-7940.
Cell 62, 793-805.
Sorger,
P. K., and Nelson,
H. C. M. (1989).
Cell 59, 807-813.
Sorger,
P. K., and Pelham,
H. R. B. (1988).
Cell 54, 855-864.
Theodorakis, N. G., Zand, D. J., Kotzbauer, P. T., Williams, Morimoto, R. I. (1989). Mol. Cell. Biol. 9, 3166-3173. Wiederrecht,
G., Seto, D., and Parker,
Xiao, H., and Lis, J. T. (1988). Xiao, H., Perisic, Zimarino,
Science
C. S. (1968).
Cell 54,841-853.
239, 1139-1142.
O., and Lis, J. T. (1991).
V., Wilson,
G. T., and
S., and Wu, C. (1990).
Cell 64, 585-593. Science
249, 546-549.
