Cell, Vol. 66, 191-197,
July 26, 1991, Copyright
0 1991 by Cell Press
Heat Shock, Stress Proteins, Chaperones, and Proteotoxicity Lawrence E. Hightower Department of Molecular and Cell Biology University of Connecticut Storrs, Connecticut 06269-3044
Cold Spring Harbor was the scene of the first international heat shock meeting, held in 1982 on the 20th anniversary year of Ferruccio Ritossa’s discovery of the heat shock response. Stimulated in part by the success of that first meeting, progress has been impressive: much of it was described at the meeting this spring at Cold Spring Harbor Laboratory, arranged by Richard Morimoto (Northwestern University) and Costa Georgopoulos (University of Utah), and entitled “Stress Proteins and The Heat Shock Response” (April 29-May2,1991). Judging from the numbers of new faces and the level of enthusiasm at this year’s meeting, the progress is bound to continue (for comprehensive reviews of the field, see Morimoto et al., 1990; Nover, 1991). Induction
Pathways
Stress Proteins and Proteotoxicity A hallmark of the heat shock response is the broad variety of stressors besides heat that induce the response: amino acid analogs, puromycin, ethanol, heavy metal ions, arsenicals, tissue explantation, infection by certain viruses, and more. The term stress proteins was introduced to recognize the more general nature of the response; glucoseregulated proteins (GRPs) are members of heat shock protein (HSP) families located in the rough endoplasmic reticulum (ER), and these are now considered as stress proteins. This was a logical extension because GRPs are induced by a variety of stressors besides glucose deprivation, including certain HSP inducers, such as heat and amino acid analogs, as well as stressors selective for GRPs, such as glycosylation inhibitors and Ca’+ ionophores. On balance, the term stress protein has been useful, but one drawback has been the perception created of a highly generalized response with no rules of engagement. Also, investigators have found utility in describing HSPs and GRPs as subsets of stress proteins, thus retaining the historical terms, particularly when studying stressors selective for each set. The pitfalls to be avoided are calling HSPs the stress proteins and the heat shock response me stress response; there are many cellular stress responses. There are rules governing the induction, and the idea that stressors that induce HSPs and GRPs share the common property of either damaging proteins directly or causing cells to synthesize aberrant proteins in the nucleocytoplasmic and ER compartments, respectively, has received considerable support from studies of both the prokaryotic and eukaryotic responses (Kozutsumi et al., 1988; Edington et al., 1989; Parsell and Sauer, 1989). At this year’s meeting, several groups using avarietyof meth-
Meeting Review
ods for introducing proteins into cells-including electroporation (Michel, Bern University; Arrigo, Claude Bernard University), transfection of genes encoding thermolabile proteins into mammalian cells (Bensaude, Ecole Normale Superieure), and microinjection into Xenopus oocytes (Mifflin, University of California, Los Angeles)-provided additional evidence that cells can distinguish between native and denatured forms of the same protein. Evidence of aggregation of denatured proteins with the 70 kd heat shock cognate protein (Hsc70) was obtained, and it was suggested that sequestering Hsc70 in this way may trigger induction. (Cognates are constitutive members of stress protein families that function in normal cellular physiology.) A popular idea is that such a sequestration could release heat shock transcription factor (HSF) from a reversible association with Hsc70, thus freeing HSF to turn on heatshockgenes. Complexescontaining both HSFand Hsc70 were found in Helacell extracts (Baler, University of Miami), adding some experimental weight to this idea. Welch (University of California, San Francisco) and Voellmy (University of Miami) are analyzing early steps in the induction pathway linked to protein synthesis in mammalian cells. In normal cells (Figure 1A) Hsc70 transiently associates with nascent polypeptide chains (Beckmann et al., 1990). However, in stressed cells (Figure IB) the association is longer lasting and could deplete Hsc70 pools, creating a deficiency in Hsc70 functions. Activation of a human hsp70 gene by a moderate heat shock is reduced by cycloheximide, probably because of a reduction in newly synthesized polypeptides, which are likely to be sensitive targets for thermal damage and which would release Hsc70 normally bound to new polypeptides into an available pool. In a structural sense, nascent polypeptides on polysomes and polypeptides unfolded for translocation across cell membranes may be viewed as transient aberrant proteins in cells. This view allows a unification of Hsp70 family function in which Hsc70 recognizes such transiently abnormal proteins in normal cells, and in stressed cells Hsc70 function is augmented by Hsp70 to handle proteins damaged directly or indirectly by stressors. The functional similarity of constitutive Hsc70 and inducible Hsp70 remains an open question. By analogy with the term genotoxicity (damage to DNA by chemical and physical agents), the term proteotoxicity should be useful to describe damage to proteins caused by chemical and physical agents. The HSPs and GRPs become subsets of stress proteins induced by proteotoxic agents. Cellular protein homeostasis is their venue. Who’s Feeling the Heat? The molecular nature of the cellular thermometer was a recurrent theme during the meeting. Gross (University of Wisconsin) summarized the evidence that Escherichia coli responds to deficiencies in functions of at least three HSPs-DnaK, DnaJ, and GrpE-by increasing the amount of the bacterial heat shock transcription factor rY2. DnaK
Cell 192
gesting the existence of two separable induction pathways. Too monolithic a view of the process is not likely to stand long. My mental image for thermal damage in cells is the potter’s kiln candles that melt at a particular temperature. The molecular “kiln candle” could be one or a class of thermolabile proteins that denature and bind up the available pools of DnaK in bacteria, and Hsc70 in eukaryotes, and recruit DnaK/Hsc70from other reversibly bound proteins such as oX2and HSF. There is evidence of cellular protein denaturation within the temperature range of induction of HSPs in both bacteria and mammalian cells (Lepock et al., 1988, 1990). Chemically damaged, analogsubstituted, and exogenously introduced denatured proteins would likely have the same effect. Alternatively, or in addition, the cellular thermometer protein and/or RNA complex itself might feel the heat directly and dissociate at heat shock temperatures. Damaged
Figure
1. Cellular
Location
and Proposed
Proteins
Roles of HSPs
(A) Chaperoning of newly synthesized proteins in normal eukaryotic cells. (6) Chaperoning of damaged proteins in stressed cells. ER, endoplasmic reticulum; Mt, mitochondrion; NC, nucleolus; Nu, nucleus; Polys, polysomes.
is the bacterial homolog of the eukaryotic Hsp70 family. The most direct way to build a thermometer would be to bind tsa2to one or more of these HSPs as negative regulators (Craig and Gross, 1991). It is also known that @ is under translational control; two cis-acting regions of the rpoH gene, encoding c 32, that are involved in translational control have been identified (Yura and Nagai, Kyoto University). RNA secondary structure predictions suggest the possibility of base pairing between these regions, which would include the translational start codon. Such a potential regulatory region combined with a regulatory protein could be a thermometer, and binding DnaK into the complex would provide a direct way to monitor heat shock protein demand. Neidhardt (University of Michigan) suggested that the ribosome may be the sensor from which signals emanate to control three interrelated stress response networks that regulate E. coli gene expression at high and low temperatures (VanBogelen and Neidhardt, 1990). He suggested that the heat shock, cold shock, and stringent networks all cooperate to allow E. coli to match protein production with growth rates over a broad temperature range. There are many possibilities for the putative signals emanating from E. coli ribosomes, including those from nascent, partially folded polypeptide chains that felt the heat and denatured. Who, then, feels the heat? Yuzawa (Kyoto University) isolated two groups of mutants in the E. coli induction pathway: one group was defective in responding to unfolded proteins as inducers, but still heat inducible, and the other group was unresponsive to either inducer, sug-
A Proliferation of Eukaryotic Heat Shock Transcription Factors Heat and other proteotoxic agents activate preexisting HSFs (reviewed by Sorger, 1991). Their DNA-binding sites (heat shock elements) are contiguous arrays of the 5 bp sequence nGAAn arranged in alternating orientations. Sorger (University of California, San Francisco) reported that the Saccharomyces cerevisiae factor scHSF has two physically separable transcriptional activation domains: the N-terminal activator, mediating the transient heat shock response, and the C-terminal activator, mediating a sustained response at elevated temperatures. scHSF binds to DNA as a trimer and multimers thereof; a region of the factor with the potential to form three-stranded coiled coils (leucine zipper motif), which may be the trimerization interface, was predicted. Wu (NIH) showed that the sequences encoding Drosophila and human HSFs contain a conserved DNA-binding domain and four leucine zipper motifs. The fourth zipper motif is absent from scHSF, and it was suggested that this zipper might allow Drosophila HSF to form homo- and heterodimers. Inactive dimers rearrange to form active hexamers upon heating. Of interest to fanciers of Hsc70-HSF complexes is the recent prediction (Lupas et al., 1991) of a coiled-coil domain in human Hsp70 (amino acids 496-542) that could provide an interaction site with zippers on HSF. Wu suggested yet another thermometer: the HSF itself. The biochemical environment of the HSF, at least in vitro, can indeed cause HSF to oligomerize, presumably through effects on protein conformation caused by heating, low pH, chaotropic agents, and detergents (Morimoto). Morimoto sounded a note of caution, based on studies using hemin and sodium salicylate as inducers, that the levels of HSF measured by in vitro assays may not correlate with the avidity of factor binding in vivo, and that DNA binding does not necessarily imply transcriptional activation. The discovery of new eukaryotic HSFs was a major topic of discussion. Morimotodescribed the cloning and characterization of murine mHSF1 and mHSF2, which respectively display inducible and constitutive binding to DNA. Schuetz (Harvard Medical School) described the isolation
Meeting 193
Review
of a second human factor (hHSF2) that appears to be different from Wu’s hHSF. Whether hHSF2 binds constitutively or inducibly to DNA is not yet clear, but it is observed to translocate from cytoplasm to nucleus upon heat shock. Westwood (NIH) revisited the intracellular localization of Drosophila dHSF, which was originally thought to be cytoplasmic, and concluded that dHSF is located in nuclei in non-heat-shocked cells, based on immunohistochemical studies. Apparently, dHSF leaks into the cytoplasm during subcellular fractionation. If Drosophila does lack a cytoplasmic counterpart to hHSF2, then this may explain a perplexing difference in the ways in which Drosophila and vertebrate cells “see” cytoplasmic proteotoxicity. For example, the intracellular accumulation of amino acid analog-substituted polypeptides leads to the induction of stress proteins in virtually all vertebrate cells but not in Drosophila. Perhaps vertebrate HSF is a cytoplasmic sentinel for proteotoxicity, whereas in Drosophila, HSF is more sensitive to damage in the nuclear compartment and the cytoplasm may be monitored differently. The current leader in the HSF extravaganza is Lycopersicon peruvianum, a tomato. Nover and Scharf (Institute of Plant Biochemistry, Halle) described work on the cloning and initial analysis of three plant HSF genes, two of which are heat inducible and therefore new HSPs as well. All share a predicted DNA-binding domain that is very similar to the comparable domain in scHSF. Several interesting examples of inhibition or attenuation of induction were presented. The flavonoid quercetin, an antioxidant, inhibited the induction of HSPs and heme oxygenase in human cells, apparently at or before activation of HSF binding (Hirayoshi, Kyoto University; Kantengwa, University Hospital, Geneva). For tobacco enthusiasts, nicotine turns out to be a coinducer of HSPs, along with either ethanol or heat (Hahn, Stanford University), while studies of natural attenuation have shown age-dependent reductions in transcriptional activation of heat shock genes in senescent human diploid lung fibroblasts (Liu, Rutgers University), and reductions in accumulation of Hsp70 mRNA in aging rodents (Holbrook, NIH). HSPs Cooperate-
Molecular
Chaperoning
One of the most exciting concepts to emerge from the study of HSPs is their role as molecular chaperones (reviewed by Ellis, 1990). Molecular chaperones are proteins that mediate the folding of other polypeptides and, in some cases, their assembly into oligomeric structures. A crucial part of the definition is that the chaperone is not part of the finished assembly. A major unifying theme that emerged from the meeting is that HSPs work together to regulate the response, to assemble/disassemble structures, and to provide a molecular shuttle service for polypeptides by chaperoning in tandem. Tracing’the Chaperoning Pathways in Eukaryotic Cells Starting with the observation that Hsc70 transiently associates with many nascent polypeptide chains in normal mammalian cells, we can begin to trace some of the poly-
peptide shuttle routes within cells (Figure 1A). Hsc70 is thought to deliver polypeptides synthesized in the cytoplasm to the rough ER in an unfolded state ready for membrane translocation. Studies of KAR2 (Grp78) mutants in yeast have shown that Grp78, located in the ER lumen, plays an essential role in the translocation process by interacting with transferred polypeptides. Grp78 may also interact with yeast SEC63, a relative of bacterial DnaJ that is required for both nuclear and ER protein translocation in yeast (Vogel, Princeton University). In mammalian cells, Grp78 (BiP) transiently associates with nascent glycoproteins and secretory proteins such as immunoglobulin polypeptides during assembly in the ER and may facilitate this process by preventing aggregation of the polypeptides and by retaining them in an assemblycompetent shape (Kaloff, University of Cologne). Grp78 has an ATPase activity that requires CaZ+, which is stored in the ER, and, like other Hsp70 family members, its release from unfolded protein substrates is accompanied by ATP hydrolysis (Hendershot, St. Jude’s Children’s Hospital). In COS cells, coexpression of heavy and light immunoglobulin chains results in the formation and secretion of heavy-light chain complexes, a reduction in levels of heavy chain-Grp78 complexes in the ER, and a reduction in levels of expression of GRPs (Lenny, St. Louis University). The paramyxoviral glycoprotein HN appropriates Grp78 for its own chaperoning needs, and the flux of HN through the ER is thought to cause induction of Grp78 during viral infection (Lamb, Northwestern University). These observations are consistent with the idea that sequestration of Grp78 bound to polypeptides is part of the signaling mechanism for induction of GRPs. Hsc70 also hands off unfolded polypeptides to mitochondrial Hsp70 (mtHsp70) via the membrane translocation system (Neupert et al., 1990; Mizzen et al., 1991). mtHsp70, in turn, may pass off at least some of these polypeptides to Hsp60/cpnlO for folding. The 10 kd chaperonin cpnl0 is a homolog of E. coli GroES and works in conjunction with Hsp60, also known as cpn60, a GroEL homolog. In the absence of Hsp60, the pathway backs up and precursor polypeptides with uncleaved presequences accumulate on the surface of mitochondria (Hallberg, Syracuse University). Another yeast mitochondrial protein, SCJl, which is homologous to SEC63 and bacterial DnaJ, is required for efficient import of polypeptides in vitro and may work with mtHsp70 (Blumberg, Princeton University). Just when it appeared safe to leave Hsc70 in the cytoplasm, Tanguay (University of Laval) reported that Drosophila Hsc70-4 is located at or near the inner mitochondrial membrane, suggesting that additional compartmentalization may underlie the proliferation of fly cognates. Chaperoning Under Stress Ironically, the mechanisms by which HSPs function in stressed cells are less clear. In Figure 1 B, thermally denatured proteins are shown as nascent chains on polysomes, in the cytoplasm, in the ER, and as part of damaged nucleoli. Both Hsp70 and Hso70 target to damaged nucleoli in heat-shocked mammalian cells, and morpho-
Cell 194
logical recovery of the nucleoli roughly correlates with the return of Hsp70 to the cytoplasm, as though repairs were completed (Pelham, 1990). This may be a variation on a normal theme, since Welch has evidence that Hsc70 is involved in preribosome assembly in nonstressed cells. We suspect that Hsp70 aids Hsc70 in salvaging denatured proteins by solubilizing them and facilitating refolding, or perhaps chaperoning them to a degradative system. We know, for example, that DnaK can “resurrect” thermally denatured RNA polymerase in vitro (Skowyra et al., 1990) and DnaK along with DnaJ and GrpE has been implicated in the renaturation of thermally denatured h repressor (Gaitanaris et al., 1990). Preliminary evidence from Lepock’s group (University of Waterloo) suggests that Hsc70 may partially refold apocytochrome c in vitro. Using Sindbis viral capsid polypeptide synthesis in vitro as a model for Hsp70-nascent polypeptide interactions, Schlesinger (Washington University) showed that several Hsp70 family proteins blocked folding of the nascent polypeptide into an autoprotease and/or blocked the cleavage site. In the ER, induction of Grp78 appears to be a consequence of sequestering Grp78 into long-lived complexes. Lee (University of Southern California) has now mapped the region of the grp78 promoter that is responsive to the accumulation of abnormal proteins in the ER and has shown that the same cis-acting regulatory elements are also responsive to a glycosylation block and a calcium ionophore. HSP Structure
and Function-Real
Beef
Creighton (EMBL) provided an overview of biochemical and biophysical approaches to studying protein folding pathways in vitro (Creighton, 1988). He raised the possibility of interactions between HSPs and a compact transition state, the “molten-globule” state, which may be a collapsed, unfolded conformation with native-like secondary structure but unstable disulfide bonds. The refolding of ribulose bisphosphate carboxylase to form a catalytically active dimer requires ATP hydrolysis, and the rate of reconstitution is accelerated about 8-fold in the presence of chaperones (Goloubinoff et al., 1989). van der Vies (Du Pont) showed that the form of this enzyme initially associated in an in vitro folding reaction with GroEL and GroES, the bacterial homologs of Hsp60 and cpnl0, has some of the structural characteristics of molten globule-like folding intermediates. Using a synthetic peptide that is largely unstructured in solution, Landry (University of Texas) showed that the peptide forms an a helix when bound to GroEL, suggesting that the peptide-binding site may accommodate protein segments as helices. Hart1 (University of Munich) presented a model emphasizing protein folding on the surfaces of GroEL (Hsp60; see Mt in Figure !A). Also describing an in vitro folding system, he showed that GroES accelerates folding of monomeric-dihydrofolate reductase but is not absolutely required, whereas both GroEL and GroES are required for rhodanese folding, with GroES possibly regulating the ATPase activity of GroEL. The basis for this new twist may lie in the innate ability of DHFR, but not rhodanese, to fold spontaneously. It was proposed that ATP-dependent release of unfolded sub-
strate proteins from GroEL is accompanied by partial folding. One of the more fascinating biological stories of the meeting involved prokaryotic endosymbionts harbored by specialized aphid fat body cells called bacteriocytes (Moriokaand Ohtaka, UniversityofTokyo). Endosymbiontssynthesize predominantly one protein, called symbionin, that is 85% identical to GroEL, the bacterial homolog of Hsp60, and can reconstitute ribulose bisphosphate carboxylase, the chloroplast Con-fixing enzyme, from its unfolded subunits in vitro. Production of symbionin decreases dramatically in isolated symbionts but is reinducible by heat, heavy metal ions, or ethanol. It was suggested that symbionin functions as a molecular chaperone supplied to the bacteriocyte by its prokaryotic passenger. A highlight of the meeting was the presentation by McKay (Stanford University) of the three-dimensional structure, at 2.2 A resolution, of the 44 kd ATPase domain of bovine Hsc70,(Flaherty et al., 1990). Use of the 44 kd fragment was necessitated by the unwillingness of the intact protein to crystallize. The ATPase fragment has a two-lobed, four-domain structure with ATP bound at the base of a deep cleft between the lobes (Figure 2). The tertiary fold of the nucleotide-binding portion resembles the functionally analogous region of hexokinase, and the three-dimensional structure of the 44 kd fragment closely resembles that of actin. The remarkable match to actin was unexpected; the amino acid sequence matches between the two proteins are essentially random. However, about two-thirds of the backbone residues in the core structures of the actin and the 44 kd fragment are positioned identically, with differences relegated mainly to the surface loops. What clues to Hsc70 function might be gleaned from these structural analogies? McKay speculated that ATP binding and hydrolysis may be involved in regulating Hsc70 oligomerization, by analogy with ATP-regulated actin polymerization. One of the striking properties of hexokinase is substrate-driven conformational changes, i.e., induced fit, and evidence was presented by Fink (University of California, Santa Cruz) that Hsc70 may undergo conformational changes upon ATP binding and interaction with unfolded proteins. Fink also showed that ADP binds to Hsc70 with 5-fold higher affinity than does ATP, suggesting that Hsc70-ADP may be the most abundant intracellular form of free cognate, and the preferred form for binding unfolded proteins. The high level of amino acid sequence identity between Hsc70 and DnaK within the ATP-binding domain promises to speed progress. DnaK mutants in which a candidate proton acceptor during ATP hydrolysis, Glu-171, is replaced by Lys, or Ala-l 74 is replaced by Thr, have altered rates of ATP hydrolysis and autophosphorylation (Kamath-Loeb, University of Wisconsin). Bovine Hsc70 has comparably positioned Glu-175 and Ala-l 78. DnaK autophosphorylation occurs on Thr-I 99, and when this residue is replaced with either Val, Ala, or Asp, the mutant DnaK binds ATP but does not autophosphorylate and ATPase activity is altered (McCarty, MIT). Interestingly, Thr-204 of Hsc70 is located in a conserved loop adjacent to the
Meeting 195
Review
phosphate moiety of bound ATP. Hsp70 family members may be capable of autophosphorylation under certain conditions, since Lee has reported phosphorylation of Grp78 and Welch hasobserved phosphorylation of mitochondrial Hsp70. In agreement with McKay’s structure of the ATPase domain, Eisenberg and Green (NIH) found only one ATP bound per Hsc70 at saturation. Using a sensitive singleburst assay for peptide-stimulated ATP hydrolysis, they reported a 1OO-fold stimulation of Hsc70 ATPase activity by clathrin baskets, and suggested that the rate-limiting step in the ATP hydrolytic cycle is product release. Additional evidence in support of this idea was given by Sadis (University of Connecticut), who showed that apocytochrome c, but not the folded holoprotein cytochrome c, stimulates ATPase activity and accelerates the exchange of ADP for ATP on bovine Hsc70. Sadis also gave secondary structure predictions for the C-terminal peptide-binding domain of Hsc70, based on circular dichroism spectroscopy and computer algorithms. There were several strong predictions for a helices, two of which were amphiphilic (amino acids 51 l-535 and 586-604). Since other proteins that bind diverse peptides, such as the human class I major histocompatibility antigen HLA-A2, also contain a helices in their peptide-binding domains, it was suggested that a channel formed between a helices could function to restrict Hsc70 binding to extended or unfolded polypeptides (Sadis et al., 1990).
Actin
Hexokinase
HSC70
ATPase
Fragment
Figure 2. Three-Dimensional Structures of Full-Size nase, and the ATPase Fragment of Bovine Hsc70
Actin,
Hexoki-
The actin structure was determined by Kabsch et al. (1990), the hexokinase structure by Fletterick et al. (1975). Schematic drawings kindly provided by D. McKay.
Steroid Hormone Receptors Hsp90 associates with a variety of steroid hormone receptors within cells, presumably protecting the receptor from inappropriate liaisons that might lead to inactivation before the steroid binds (Cadepond, INSERM). The untransformed steroid hormone receptor complex provides another example where multiple HSPs appear to cooperate, in this case by chaperoning the receptor until steroid hormones interact with it. Toft (Mayo Foundation) described an exciting approach to studying the interactions of Hsp90 and Hsp70 with the chicken progesterone receptor. He showed that purified receptor, stripped of HSPs and bound to an antibody affinity resin, will rebind Hsp90 and Hsp70 in a reticulocyte lysate. He has begun a functional-domain analysis of Hsp90, which already indicates that internal deletions betweeen residues 381 and 708 dramatically reduce receptor binding. Hsp90 has already been implicated in interactions with transcription factors, i.e., steroid hormone receptors, and is a very abundant protein. Thus, there is ample Hsp90 to accommodate other transcription factors, inclucing HSF, in preinduction complexes-perhaps our Hsc70-HSF complex will turn out to be HspSO-Hsc70-HSF. Another glucocorticoid receptor-associated protein (59 kd) was shown to be a novel, low abundance HSP (Sanchez, Medical College of Ohio), so the hypothetical complex could grow larger yet. Bacterial and Phage Models Studies of the assembly of preprimosomes for initiation of phage h DNA synthesis by DnaK and other E. coli HSPs
Cell 196
(GrpE/DnaJ) have yielded the most comprehensive view to date of the roles of HSPs in protein assembly, and provide another example of cooperation among HSPs (reviewed by Georgopoulos et al., 1990). Steps in h DNA replication include the following: the h0 initiator protein binds specifically to or&, the hP initiator protein complexes with the E. coli DnaB helicase, and the hP-DnaB complex becomes localized to orih through an ho-hP interaction. The overall role of HSPs in DNA replication is the dissociation of the protein complex located at orih and the sequestration of the hP protein, allowing activation of DnaB and the initiation of replication. Zylicz (University of Gdansk) showed that DnaJ and GrpE stimulate the ATPase activity of DnaK, modulate conformational changes in DnaK triggered by ATP hydrolysis, and modulate the stability of the hP-DnaK complex. DnaJ stimulates the rate of ATP hydrolysis by DnaK, and GrpE increases the rate of ADP release from DnaK (K. Liberek, University of Utah). Additional evidence was presented that a DnaJ-DnaK-ATP interaction is needed to release protein P from preinitiation complexes (Hoffmann, University of California, Berkeley). Using a phage Pi-based model of DNA replication, Wickner (NIH) showed that DnaK and DnaJ activate the RepA replication initiation protein by converting RepA dimers to monomers. This unexpected finding is particularly exciting because it suggests possible analogies with the c-Jun and c-Fos transcription factors, whose activities appear to be modulated by conversion of homodimers to more potent heterodimers. The DnaWHsp70 proteins may facilitate such conversions, allowing cells to respond quickly to new environmental signals. Acquired Thermotolerance and Thermoresistance When either embryos or cultured cells are subjected to a mild heat stress or to certain chemical stressors, they acquire the capacity to survive what would normally be lethal exposures (reviewed by Hahn and Li, 1990; Petersen, 1990). Virtually every major class of HSPs has now been implicated in this acquired thermotolerance. Lindquist (University of Chicago) showed that Hsp70 plays a major role in thermotolerance in Drosophila based on mutational analyses, overexpression of Hsp70, and inhibition of Hsp70 expression using antisense RNA. She also reported on studies of yeast deletion mutants indicating that Hspl04 plays a major role in induction of thermotolerance. Interestingly, the mutants still retain a transient capacity to tolerate heat, an indication of the multifaceted nature of this response. Li (Sloan-Kettering Cancer Center) used a murine retroviral vector to express constitutively a human hsp70 gene in rodent cells. Stable transformants displayed increased levels of resistance to thermal killing that correlated with the amounts of Hsp70 present. In other studies, thermotolerant mammalian cells accumulated less protein damage than unprotected cells, and Hsp70 was implicated in protection/repair of protein damage in thermotolerant cells (Kim, Stanford University). Evidence was presented that bovine Hsc70, added exogenously to cultured arterial cells and injected into rat retina, may prolong arterial cell survival in culture and protect retinal cells from light damage (Tytell, Wake Forest University).
Induction of Hsp27 alone, using an inducible expression vector driven by a metallothionein promoter, rapidly conferred protection against thermal killing on mouse cells in culture (Lavoie, University of Laval). Phosphorylation of preexisting Hsp27 is an early event in mammalian cellular responses to stress and was suggested as a possible signal for development of thermotolerance (Landry, University of Laval). The a-crystallins of vertebrate lenses belong to the same superfamily as small HSPs like Hsp27. It has now been shown that aB-crystallin is an HSP and the promoter of its gene contains a heat shock element. Remarkably, overexpression of aB-crystallin in mammalian cells leads to the acquisition of thermotolerance (Chepelinsky, NIH; Klemenz, University of Zurich). Then, there are those disturbing instances when HSPs and thermotolerance come apart. For example, yeast cells with a mutation in their HSF gene are defective for growth and mitochondrial protein import at elevated temperature, but are still able to acquire thermotolerance (Smith, University of California, San Diego). A mouse lymphoma B cell line acquires thermotolerance in culture and when grown as an ascites tumor, but manages to induce HSPs only in the latter case (Anderson, Peter MacCallum Cancer Institute). Conversely, induction of HSPs in E. coli without a temperature shift is insufficient to induce thermotolerante (VanBogelen, University of Michigan). Obviously, understanding thermotolerance remains for the future. Fishing for Genetic Variation in Streams and Culture Dishes Another future challenge will be to apply our knowledge of stress responses to the natural environment. For example, the impact of the Big Heat Shock, i.e., rapid global warming, and other forms of pollution may depend on how much genetic variation in resistance mechanisms is available in natural populations. This is largely unknown. Clonal reproduction among certain hybrid species of the Sonoran Desert topminnow Poeciliopsis allows amplification and study of naturally occurring combinations of genetic alleles in the laboratory. This has led to evaluation of the deploy ment of genetic resistance mechanisms such as the heat shock response in natural populations (Schultz, 1989). Hightower showed that these desert fish have a highly exaggerated heat shock response, presumably part of their adaptation, and different species have elaborated different isoforms of the Hsp70 and Hsp20-30 families that may contribute to observed differences in thermal resistance among biotypes. Several groups described what may be cell culture versions of the experiments of nature described above. Anderson and Hahn (Stanford University) selected heatresistant variants of mammalian cells that express elevated levels of the major HSPs and have a new Hsp70 family isoform. Laszlo (Washington University) described heat-sensitive variants unable to develop thermotolerance and unable to recover rapidly from thermal damage, and heat-resistant variants that overexpressed Hsp70 and Hsp27. These data suggest that thermal adaptation at the cellular level is associated with changes in the heat shock response. Gupta (McMaster University) showed that mu-
Meeting 197
Review
tants of a mammalian cell line selected for resistance to microtubule inhibitors contain new isoforms of Hsp60 and Hsc70, suggesting that they may interact with tubulins during microtubule assembly (Gupta, 1990). These studies taken together raise the possibility that adaptation to stress by cells in culture and organisms in the environment may require changes in chaperones. The exploration of these changes may yield new insights into the evolution of protein folding and assembly pathways. Acknowledgments I wish to thank Costa Georgopoulos, David McKay, Rick Morimoto, Seth Sadis, and Bill Welch for reading and commenting on the manuscript. Unfortunately, there were many more interesting presentations than could be fit into this review. References Beckmann, Ft. P., Mizzen, L. A., and Welch, W. J. (1990). Interaction of Hsp70 with newly synthesized proteins: implications for protein folding and assembly. Science 248, 850-854. Craig, E. A., and Gross, C. A. (1991). Is hsp70 the cellular ter? Trends Biochem. Sci. 76, 135-139.
thermome-
Creighton, T. E. (1988). Toward a better understanding of protein ing pathways. Proc. Natl. Acad. Sci. USA 85, 5082-5086.
fold-
Edington, B. V., Whelan, S. A., and Hightower, L. E. (1989). Inhibition of heat shock (stress) protein induction by deuterium oxide and glycerol: additional support for the abnormal protein hypothesis of induction. J. Cell. Physiol. 739, 219-228. Ellis, Ft. J. (1990).
Molecular
chaperones.
Semin.
Cell Biol. 7, I-72.
Flaherty, K. M., DeLuca-Flaherty, C., and McKay, D. B. (1990). Threedimensional structure of the ATPase fragment of a 70K heat-shock cognate protein. Nature 346, 623-628.
of bacteria by differential scanning denaturation in situ to maximum phys. Acta 7055, 19-26.
calorimetry: relationship of protein growth temperature. Biochim. Bio-
Lupas, A., Van Dyke, M., and Stock, J. (1991). from protein sequences. Science 252, 1162-l
Predicting 164.
coiled coils
Mizzen, L. A., Kabiling, A. N., and Welch, W. J. (1991). The two mammamlian mitochondrial stress proteins, grp75 and hsp58, transiently interact with newly synthesized mitochondrial proteins. Cell Reg. 2, 165-179. 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). Neupert, W., Hartl, F.-U., Craig, E. A., and Pfanner, polypeptides cross the mitochondrial membranes? Nover, L. (1991). CRC Press).
The
Heat Shock
Response
(Boca
N. (1990). How do Cell 63, 447-450. Raton,
Florida:
Parsell, D. A., and Sauer, R. T. (1989). Induction of a heat shocklike response by unfolded protein in Eschericbia co/i: dependence on protein level not protein degradation. Genes Dev. 3, 1226-1232. Pelham, H. R. B. (1990). Functions of the hsp70 protein family: an overview. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 287-299. Petersen, N. S. (1990). Effects of heat and chemical stress on development. In Advances in Genetics: Genomic Responses to Environmental Stress, J. G. Scandalios, ed. (New York: Alan R. Liss), pp. 275-296. Sadis, S., Raghavendra, K., and Hightower, L. E. (1990). Secondary structure of the mammalian 70-kilodalton heat shock cognate protein analyzed by circular dichroism spectroscopy and secondary structure prediction. Biochemistry 29, 8199-8206. Schultz, R. J. (1989). Origins and relationships of unisexual poeciliids. In Evolution and Ecology of Poeciliid Fishes, G. K. Meffe and F. F. Snelson, Jr., eds. (Englewood Cliffs, New Jersey: Prentice-Hall), pp. 69-87.
Fletterick, Ft. J., Bates, D. J., and Steitz, T. A. (1975). The structure of a yeast hexokinase monomer and its complexes with substrates at 2.7 A resolution. Proc. Natl. Acad. Sci. USA 72, 38-42.
Skowyra, D., Georgopoulos, C., and Zylicz, M. (1990). The E. coli dnaK gene product, the hsp70 homolog, can reactivate heat-inactivated RNA polymerase in an ATP hydrolysis-dependent manner. Cell 62, 939-944.
Gaitanaris, G. A., Papavassiliou, A. G., Rubock, P., Silverstein, S. J., and Gottesman, M. E. (1990). Renaturation of denatured h repressor requires heat shock proteins. Cell 67, 1013-1020.
Sorger, P. K. (1991). Cell 65, 363-366.
Georgopoulus, C., Ang, D., Liberek, K., and Zylicz, M. (1990). Properties of the Escherichia co/i HSPs and their role in bacteriophage I growth. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 191-221. Goloubinoff, P., Christeller, J. T., and Gatenby, A. A. (1989). Reconstitution of active dimeric ribulose bisphosphate carboxylase from an unfolded state depends on two chaperonin proteins and Mg-ATP. Nature 342, 884-889. Gupta, R. S. (1990). the in viva assembly 418.
Mitochondria, molecular chaperone of microtubules. Trends Biochem.
proteins and Sci. 75, 415-
Hahn, G. M., and Li, G. C. (1990). Thermotolerance, thermoresistance, and thermosensitization. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 79-100. Kabsch, W., Mannherz, (1990). Atomic structure 37-44.
H. G., Suck, D., Pai, E. F., and Holmes, K. C. of the actin:DNase I complex. Nature 347,
Kozutsumi, Y., Segal, M., Normington, K., Gething, M.J., and Sambrook, J. (1988). The presence of malfolded proteins in the endoplasmic reticulum signals the induction of glucose-regulated proteins. Nature 332, 462-464. Lepock, J. R., Frey, H. E., Rodahl, A. M., and Kruuv, J. (1988). Thermal analysis of CHL V79 cells using differential scanning calorimetry: implications for hyperthermic cell killing and the heat shock response. J. Cell. Physiol. 737, 14-24. Lepock,
J. R., Frey, H. E., and Inniss,
W. E. (1990). Thermal
analysis
Heat shock
factor
and the heat shock
response.
VanBogelen, R. A., and Neidhardt, F. C. (1990). Ribosomes as sensors of heat and cold shock in Escherichia co/i. Proc. Natl. Acad. Sci. USA 87. 5589-5593. Note
Added
Induction changes
in Proof
of HSPs can also be modulated in Drosophila in glutamine concentrations (Sanders, UMDNJ).
cells
by
July 26, 1991, Copyright
0 1991 by Cell Press
Heat Shock, Stress Proteins, Chaperones, and Proteotoxicity Lawrence E. Hightower Department of Molecular and Cell Biology University of Connecticut Storrs, Connecticut 06269-3044
Cold Spring Harbor was the scene of the first international heat shock meeting, held in 1982 on the 20th anniversary year of Ferruccio Ritossa’s discovery of the heat shock response. Stimulated in part by the success of that first meeting, progress has been impressive: much of it was described at the meeting this spring at Cold Spring Harbor Laboratory, arranged by Richard Morimoto (Northwestern University) and Costa Georgopoulos (University of Utah), and entitled “Stress Proteins and The Heat Shock Response” (April 29-May2,1991). Judging from the numbers of new faces and the level of enthusiasm at this year’s meeting, the progress is bound to continue (for comprehensive reviews of the field, see Morimoto et al., 1990; Nover, 1991). Induction
Pathways
Stress Proteins and Proteotoxicity A hallmark of the heat shock response is the broad variety of stressors besides heat that induce the response: amino acid analogs, puromycin, ethanol, heavy metal ions, arsenicals, tissue explantation, infection by certain viruses, and more. The term stress proteins was introduced to recognize the more general nature of the response; glucoseregulated proteins (GRPs) are members of heat shock protein (HSP) families located in the rough endoplasmic reticulum (ER), and these are now considered as stress proteins. This was a logical extension because GRPs are induced by a variety of stressors besides glucose deprivation, including certain HSP inducers, such as heat and amino acid analogs, as well as stressors selective for GRPs, such as glycosylation inhibitors and Ca’+ ionophores. On balance, the term stress protein has been useful, but one drawback has been the perception created of a highly generalized response with no rules of engagement. Also, investigators have found utility in describing HSPs and GRPs as subsets of stress proteins, thus retaining the historical terms, particularly when studying stressors selective for each set. The pitfalls to be avoided are calling HSPs the stress proteins and the heat shock response me stress response; there are many cellular stress responses. There are rules governing the induction, and the idea that stressors that induce HSPs and GRPs share the common property of either damaging proteins directly or causing cells to synthesize aberrant proteins in the nucleocytoplasmic and ER compartments, respectively, has received considerable support from studies of both the prokaryotic and eukaryotic responses (Kozutsumi et al., 1988; Edington et al., 1989; Parsell and Sauer, 1989). At this year’s meeting, several groups using avarietyof meth-
Meeting Review
ods for introducing proteins into cells-including electroporation (Michel, Bern University; Arrigo, Claude Bernard University), transfection of genes encoding thermolabile proteins into mammalian cells (Bensaude, Ecole Normale Superieure), and microinjection into Xenopus oocytes (Mifflin, University of California, Los Angeles)-provided additional evidence that cells can distinguish between native and denatured forms of the same protein. Evidence of aggregation of denatured proteins with the 70 kd heat shock cognate protein (Hsc70) was obtained, and it was suggested that sequestering Hsc70 in this way may trigger induction. (Cognates are constitutive members of stress protein families that function in normal cellular physiology.) A popular idea is that such a sequestration could release heat shock transcription factor (HSF) from a reversible association with Hsc70, thus freeing HSF to turn on heatshockgenes. Complexescontaining both HSFand Hsc70 were found in Helacell extracts (Baler, University of Miami), adding some experimental weight to this idea. Welch (University of California, San Francisco) and Voellmy (University of Miami) are analyzing early steps in the induction pathway linked to protein synthesis in mammalian cells. In normal cells (Figure 1A) Hsc70 transiently associates with nascent polypeptide chains (Beckmann et al., 1990). However, in stressed cells (Figure IB) the association is longer lasting and could deplete Hsc70 pools, creating a deficiency in Hsc70 functions. Activation of a human hsp70 gene by a moderate heat shock is reduced by cycloheximide, probably because of a reduction in newly synthesized polypeptides, which are likely to be sensitive targets for thermal damage and which would release Hsc70 normally bound to new polypeptides into an available pool. In a structural sense, nascent polypeptides on polysomes and polypeptides unfolded for translocation across cell membranes may be viewed as transient aberrant proteins in cells. This view allows a unification of Hsp70 family function in which Hsc70 recognizes such transiently abnormal proteins in normal cells, and in stressed cells Hsc70 function is augmented by Hsp70 to handle proteins damaged directly or indirectly by stressors. The functional similarity of constitutive Hsc70 and inducible Hsp70 remains an open question. By analogy with the term genotoxicity (damage to DNA by chemical and physical agents), the term proteotoxicity should be useful to describe damage to proteins caused by chemical and physical agents. The HSPs and GRPs become subsets of stress proteins induced by proteotoxic agents. Cellular protein homeostasis is their venue. Who’s Feeling the Heat? The molecular nature of the cellular thermometer was a recurrent theme during the meeting. Gross (University of Wisconsin) summarized the evidence that Escherichia coli responds to deficiencies in functions of at least three HSPs-DnaK, DnaJ, and GrpE-by increasing the amount of the bacterial heat shock transcription factor rY2. DnaK
Cell 192
gesting the existence of two separable induction pathways. Too monolithic a view of the process is not likely to stand long. My mental image for thermal damage in cells is the potter’s kiln candles that melt at a particular temperature. The molecular “kiln candle” could be one or a class of thermolabile proteins that denature and bind up the available pools of DnaK in bacteria, and Hsc70 in eukaryotes, and recruit DnaK/Hsc70from other reversibly bound proteins such as oX2and HSF. There is evidence of cellular protein denaturation within the temperature range of induction of HSPs in both bacteria and mammalian cells (Lepock et al., 1988, 1990). Chemically damaged, analogsubstituted, and exogenously introduced denatured proteins would likely have the same effect. Alternatively, or in addition, the cellular thermometer protein and/or RNA complex itself might feel the heat directly and dissociate at heat shock temperatures. Damaged
Figure
1. Cellular
Location
and Proposed
Proteins
Roles of HSPs
(A) Chaperoning of newly synthesized proteins in normal eukaryotic cells. (6) Chaperoning of damaged proteins in stressed cells. ER, endoplasmic reticulum; Mt, mitochondrion; NC, nucleolus; Nu, nucleus; Polys, polysomes.
is the bacterial homolog of the eukaryotic Hsp70 family. The most direct way to build a thermometer would be to bind tsa2to one or more of these HSPs as negative regulators (Craig and Gross, 1991). It is also known that @ is under translational control; two cis-acting regions of the rpoH gene, encoding c 32, that are involved in translational control have been identified (Yura and Nagai, Kyoto University). RNA secondary structure predictions suggest the possibility of base pairing between these regions, which would include the translational start codon. Such a potential regulatory region combined with a regulatory protein could be a thermometer, and binding DnaK into the complex would provide a direct way to monitor heat shock protein demand. Neidhardt (University of Michigan) suggested that the ribosome may be the sensor from which signals emanate to control three interrelated stress response networks that regulate E. coli gene expression at high and low temperatures (VanBogelen and Neidhardt, 1990). He suggested that the heat shock, cold shock, and stringent networks all cooperate to allow E. coli to match protein production with growth rates over a broad temperature range. There are many possibilities for the putative signals emanating from E. coli ribosomes, including those from nascent, partially folded polypeptide chains that felt the heat and denatured. Who, then, feels the heat? Yuzawa (Kyoto University) isolated two groups of mutants in the E. coli induction pathway: one group was defective in responding to unfolded proteins as inducers, but still heat inducible, and the other group was unresponsive to either inducer, sug-
A Proliferation of Eukaryotic Heat Shock Transcription Factors Heat and other proteotoxic agents activate preexisting HSFs (reviewed by Sorger, 1991). Their DNA-binding sites (heat shock elements) are contiguous arrays of the 5 bp sequence nGAAn arranged in alternating orientations. Sorger (University of California, San Francisco) reported that the Saccharomyces cerevisiae factor scHSF has two physically separable transcriptional activation domains: the N-terminal activator, mediating the transient heat shock response, and the C-terminal activator, mediating a sustained response at elevated temperatures. scHSF binds to DNA as a trimer and multimers thereof; a region of the factor with the potential to form three-stranded coiled coils (leucine zipper motif), which may be the trimerization interface, was predicted. Wu (NIH) showed that the sequences encoding Drosophila and human HSFs contain a conserved DNA-binding domain and four leucine zipper motifs. The fourth zipper motif is absent from scHSF, and it was suggested that this zipper might allow Drosophila HSF to form homo- and heterodimers. Inactive dimers rearrange to form active hexamers upon heating. Of interest to fanciers of Hsc70-HSF complexes is the recent prediction (Lupas et al., 1991) of a coiled-coil domain in human Hsp70 (amino acids 496-542) that could provide an interaction site with zippers on HSF. Wu suggested yet another thermometer: the HSF itself. The biochemical environment of the HSF, at least in vitro, can indeed cause HSF to oligomerize, presumably through effects on protein conformation caused by heating, low pH, chaotropic agents, and detergents (Morimoto). Morimoto sounded a note of caution, based on studies using hemin and sodium salicylate as inducers, that the levels of HSF measured by in vitro assays may not correlate with the avidity of factor binding in vivo, and that DNA binding does not necessarily imply transcriptional activation. The discovery of new eukaryotic HSFs was a major topic of discussion. Morimotodescribed the cloning and characterization of murine mHSF1 and mHSF2, which respectively display inducible and constitutive binding to DNA. Schuetz (Harvard Medical School) described the isolation
Meeting 193
Review
of a second human factor (hHSF2) that appears to be different from Wu’s hHSF. Whether hHSF2 binds constitutively or inducibly to DNA is not yet clear, but it is observed to translocate from cytoplasm to nucleus upon heat shock. Westwood (NIH) revisited the intracellular localization of Drosophila dHSF, which was originally thought to be cytoplasmic, and concluded that dHSF is located in nuclei in non-heat-shocked cells, based on immunohistochemical studies. Apparently, dHSF leaks into the cytoplasm during subcellular fractionation. If Drosophila does lack a cytoplasmic counterpart to hHSF2, then this may explain a perplexing difference in the ways in which Drosophila and vertebrate cells “see” cytoplasmic proteotoxicity. For example, the intracellular accumulation of amino acid analog-substituted polypeptides leads to the induction of stress proteins in virtually all vertebrate cells but not in Drosophila. Perhaps vertebrate HSF is a cytoplasmic sentinel for proteotoxicity, whereas in Drosophila, HSF is more sensitive to damage in the nuclear compartment and the cytoplasm may be monitored differently. The current leader in the HSF extravaganza is Lycopersicon peruvianum, a tomato. Nover and Scharf (Institute of Plant Biochemistry, Halle) described work on the cloning and initial analysis of three plant HSF genes, two of which are heat inducible and therefore new HSPs as well. All share a predicted DNA-binding domain that is very similar to the comparable domain in scHSF. Several interesting examples of inhibition or attenuation of induction were presented. The flavonoid quercetin, an antioxidant, inhibited the induction of HSPs and heme oxygenase in human cells, apparently at or before activation of HSF binding (Hirayoshi, Kyoto University; Kantengwa, University Hospital, Geneva). For tobacco enthusiasts, nicotine turns out to be a coinducer of HSPs, along with either ethanol or heat (Hahn, Stanford University), while studies of natural attenuation have shown age-dependent reductions in transcriptional activation of heat shock genes in senescent human diploid lung fibroblasts (Liu, Rutgers University), and reductions in accumulation of Hsp70 mRNA in aging rodents (Holbrook, NIH). HSPs Cooperate-
Molecular
Chaperoning
One of the most exciting concepts to emerge from the study of HSPs is their role as molecular chaperones (reviewed by Ellis, 1990). Molecular chaperones are proteins that mediate the folding of other polypeptides and, in some cases, their assembly into oligomeric structures. A crucial part of the definition is that the chaperone is not part of the finished assembly. A major unifying theme that emerged from the meeting is that HSPs work together to regulate the response, to assemble/disassemble structures, and to provide a molecular shuttle service for polypeptides by chaperoning in tandem. Tracing’the Chaperoning Pathways in Eukaryotic Cells Starting with the observation that Hsc70 transiently associates with many nascent polypeptide chains in normal mammalian cells, we can begin to trace some of the poly-
peptide shuttle routes within cells (Figure 1A). Hsc70 is thought to deliver polypeptides synthesized in the cytoplasm to the rough ER in an unfolded state ready for membrane translocation. Studies of KAR2 (Grp78) mutants in yeast have shown that Grp78, located in the ER lumen, plays an essential role in the translocation process by interacting with transferred polypeptides. Grp78 may also interact with yeast SEC63, a relative of bacterial DnaJ that is required for both nuclear and ER protein translocation in yeast (Vogel, Princeton University). In mammalian cells, Grp78 (BiP) transiently associates with nascent glycoproteins and secretory proteins such as immunoglobulin polypeptides during assembly in the ER and may facilitate this process by preventing aggregation of the polypeptides and by retaining them in an assemblycompetent shape (Kaloff, University of Cologne). Grp78 has an ATPase activity that requires CaZ+, which is stored in the ER, and, like other Hsp70 family members, its release from unfolded protein substrates is accompanied by ATP hydrolysis (Hendershot, St. Jude’s Children’s Hospital). In COS cells, coexpression of heavy and light immunoglobulin chains results in the formation and secretion of heavy-light chain complexes, a reduction in levels of heavy chain-Grp78 complexes in the ER, and a reduction in levels of expression of GRPs (Lenny, St. Louis University). The paramyxoviral glycoprotein HN appropriates Grp78 for its own chaperoning needs, and the flux of HN through the ER is thought to cause induction of Grp78 during viral infection (Lamb, Northwestern University). These observations are consistent with the idea that sequestration of Grp78 bound to polypeptides is part of the signaling mechanism for induction of GRPs. Hsc70 also hands off unfolded polypeptides to mitochondrial Hsp70 (mtHsp70) via the membrane translocation system (Neupert et al., 1990; Mizzen et al., 1991). mtHsp70, in turn, may pass off at least some of these polypeptides to Hsp60/cpnlO for folding. The 10 kd chaperonin cpnl0 is a homolog of E. coli GroES and works in conjunction with Hsp60, also known as cpn60, a GroEL homolog. In the absence of Hsp60, the pathway backs up and precursor polypeptides with uncleaved presequences accumulate on the surface of mitochondria (Hallberg, Syracuse University). Another yeast mitochondrial protein, SCJl, which is homologous to SEC63 and bacterial DnaJ, is required for efficient import of polypeptides in vitro and may work with mtHsp70 (Blumberg, Princeton University). Just when it appeared safe to leave Hsc70 in the cytoplasm, Tanguay (University of Laval) reported that Drosophila Hsc70-4 is located at or near the inner mitochondrial membrane, suggesting that additional compartmentalization may underlie the proliferation of fly cognates. Chaperoning Under Stress Ironically, the mechanisms by which HSPs function in stressed cells are less clear. In Figure 1 B, thermally denatured proteins are shown as nascent chains on polysomes, in the cytoplasm, in the ER, and as part of damaged nucleoli. Both Hsp70 and Hso70 target to damaged nucleoli in heat-shocked mammalian cells, and morpho-
Cell 194
logical recovery of the nucleoli roughly correlates with the return of Hsp70 to the cytoplasm, as though repairs were completed (Pelham, 1990). This may be a variation on a normal theme, since Welch has evidence that Hsc70 is involved in preribosome assembly in nonstressed cells. We suspect that Hsp70 aids Hsc70 in salvaging denatured proteins by solubilizing them and facilitating refolding, or perhaps chaperoning them to a degradative system. We know, for example, that DnaK can “resurrect” thermally denatured RNA polymerase in vitro (Skowyra et al., 1990) and DnaK along with DnaJ and GrpE has been implicated in the renaturation of thermally denatured h repressor (Gaitanaris et al., 1990). Preliminary evidence from Lepock’s group (University of Waterloo) suggests that Hsc70 may partially refold apocytochrome c in vitro. Using Sindbis viral capsid polypeptide synthesis in vitro as a model for Hsp70-nascent polypeptide interactions, Schlesinger (Washington University) showed that several Hsp70 family proteins blocked folding of the nascent polypeptide into an autoprotease and/or blocked the cleavage site. In the ER, induction of Grp78 appears to be a consequence of sequestering Grp78 into long-lived complexes. Lee (University of Southern California) has now mapped the region of the grp78 promoter that is responsive to the accumulation of abnormal proteins in the ER and has shown that the same cis-acting regulatory elements are also responsive to a glycosylation block and a calcium ionophore. HSP Structure
and Function-Real
Beef
Creighton (EMBL) provided an overview of biochemical and biophysical approaches to studying protein folding pathways in vitro (Creighton, 1988). He raised the possibility of interactions between HSPs and a compact transition state, the “molten-globule” state, which may be a collapsed, unfolded conformation with native-like secondary structure but unstable disulfide bonds. The refolding of ribulose bisphosphate carboxylase to form a catalytically active dimer requires ATP hydrolysis, and the rate of reconstitution is accelerated about 8-fold in the presence of chaperones (Goloubinoff et al., 1989). van der Vies (Du Pont) showed that the form of this enzyme initially associated in an in vitro folding reaction with GroEL and GroES, the bacterial homologs of Hsp60 and cpnl0, has some of the structural characteristics of molten globule-like folding intermediates. Using a synthetic peptide that is largely unstructured in solution, Landry (University of Texas) showed that the peptide forms an a helix when bound to GroEL, suggesting that the peptide-binding site may accommodate protein segments as helices. Hart1 (University of Munich) presented a model emphasizing protein folding on the surfaces of GroEL (Hsp60; see Mt in Figure !A). Also describing an in vitro folding system, he showed that GroES accelerates folding of monomeric-dihydrofolate reductase but is not absolutely required, whereas both GroEL and GroES are required for rhodanese folding, with GroES possibly regulating the ATPase activity of GroEL. The basis for this new twist may lie in the innate ability of DHFR, but not rhodanese, to fold spontaneously. It was proposed that ATP-dependent release of unfolded sub-
strate proteins from GroEL is accompanied by partial folding. One of the more fascinating biological stories of the meeting involved prokaryotic endosymbionts harbored by specialized aphid fat body cells called bacteriocytes (Moriokaand Ohtaka, UniversityofTokyo). Endosymbiontssynthesize predominantly one protein, called symbionin, that is 85% identical to GroEL, the bacterial homolog of Hsp60, and can reconstitute ribulose bisphosphate carboxylase, the chloroplast Con-fixing enzyme, from its unfolded subunits in vitro. Production of symbionin decreases dramatically in isolated symbionts but is reinducible by heat, heavy metal ions, or ethanol. It was suggested that symbionin functions as a molecular chaperone supplied to the bacteriocyte by its prokaryotic passenger. A highlight of the meeting was the presentation by McKay (Stanford University) of the three-dimensional structure, at 2.2 A resolution, of the 44 kd ATPase domain of bovine Hsc70,(Flaherty et al., 1990). Use of the 44 kd fragment was necessitated by the unwillingness of the intact protein to crystallize. The ATPase fragment has a two-lobed, four-domain structure with ATP bound at the base of a deep cleft between the lobes (Figure 2). The tertiary fold of the nucleotide-binding portion resembles the functionally analogous region of hexokinase, and the three-dimensional structure of the 44 kd fragment closely resembles that of actin. The remarkable match to actin was unexpected; the amino acid sequence matches between the two proteins are essentially random. However, about two-thirds of the backbone residues in the core structures of the actin and the 44 kd fragment are positioned identically, with differences relegated mainly to the surface loops. What clues to Hsc70 function might be gleaned from these structural analogies? McKay speculated that ATP binding and hydrolysis may be involved in regulating Hsc70 oligomerization, by analogy with ATP-regulated actin polymerization. One of the striking properties of hexokinase is substrate-driven conformational changes, i.e., induced fit, and evidence was presented by Fink (University of California, Santa Cruz) that Hsc70 may undergo conformational changes upon ATP binding and interaction with unfolded proteins. Fink also showed that ADP binds to Hsc70 with 5-fold higher affinity than does ATP, suggesting that Hsc70-ADP may be the most abundant intracellular form of free cognate, and the preferred form for binding unfolded proteins. The high level of amino acid sequence identity between Hsc70 and DnaK within the ATP-binding domain promises to speed progress. DnaK mutants in which a candidate proton acceptor during ATP hydrolysis, Glu-171, is replaced by Lys, or Ala-l 74 is replaced by Thr, have altered rates of ATP hydrolysis and autophosphorylation (Kamath-Loeb, University of Wisconsin). Bovine Hsc70 has comparably positioned Glu-175 and Ala-l 78. DnaK autophosphorylation occurs on Thr-I 99, and when this residue is replaced with either Val, Ala, or Asp, the mutant DnaK binds ATP but does not autophosphorylate and ATPase activity is altered (McCarty, MIT). Interestingly, Thr-204 of Hsc70 is located in a conserved loop adjacent to the
Meeting 195
Review
phosphate moiety of bound ATP. Hsp70 family members may be capable of autophosphorylation under certain conditions, since Lee has reported phosphorylation of Grp78 and Welch hasobserved phosphorylation of mitochondrial Hsp70. In agreement with McKay’s structure of the ATPase domain, Eisenberg and Green (NIH) found only one ATP bound per Hsc70 at saturation. Using a sensitive singleburst assay for peptide-stimulated ATP hydrolysis, they reported a 1OO-fold stimulation of Hsc70 ATPase activity by clathrin baskets, and suggested that the rate-limiting step in the ATP hydrolytic cycle is product release. Additional evidence in support of this idea was given by Sadis (University of Connecticut), who showed that apocytochrome c, but not the folded holoprotein cytochrome c, stimulates ATPase activity and accelerates the exchange of ADP for ATP on bovine Hsc70. Sadis also gave secondary structure predictions for the C-terminal peptide-binding domain of Hsc70, based on circular dichroism spectroscopy and computer algorithms. There were several strong predictions for a helices, two of which were amphiphilic (amino acids 51 l-535 and 586-604). Since other proteins that bind diverse peptides, such as the human class I major histocompatibility antigen HLA-A2, also contain a helices in their peptide-binding domains, it was suggested that a channel formed between a helices could function to restrict Hsc70 binding to extended or unfolded polypeptides (Sadis et al., 1990).
Actin
Hexokinase
HSC70
ATPase
Fragment
Figure 2. Three-Dimensional Structures of Full-Size nase, and the ATPase Fragment of Bovine Hsc70
Actin,
Hexoki-
The actin structure was determined by Kabsch et al. (1990), the hexokinase structure by Fletterick et al. (1975). Schematic drawings kindly provided by D. McKay.
Steroid Hormone Receptors Hsp90 associates with a variety of steroid hormone receptors within cells, presumably protecting the receptor from inappropriate liaisons that might lead to inactivation before the steroid binds (Cadepond, INSERM). The untransformed steroid hormone receptor complex provides another example where multiple HSPs appear to cooperate, in this case by chaperoning the receptor until steroid hormones interact with it. Toft (Mayo Foundation) described an exciting approach to studying the interactions of Hsp90 and Hsp70 with the chicken progesterone receptor. He showed that purified receptor, stripped of HSPs and bound to an antibody affinity resin, will rebind Hsp90 and Hsp70 in a reticulocyte lysate. He has begun a functional-domain analysis of Hsp90, which already indicates that internal deletions betweeen residues 381 and 708 dramatically reduce receptor binding. Hsp90 has already been implicated in interactions with transcription factors, i.e., steroid hormone receptors, and is a very abundant protein. Thus, there is ample Hsp90 to accommodate other transcription factors, inclucing HSF, in preinduction complexes-perhaps our Hsc70-HSF complex will turn out to be HspSO-Hsc70-HSF. Another glucocorticoid receptor-associated protein (59 kd) was shown to be a novel, low abundance HSP (Sanchez, Medical College of Ohio), so the hypothetical complex could grow larger yet. Bacterial and Phage Models Studies of the assembly of preprimosomes for initiation of phage h DNA synthesis by DnaK and other E. coli HSPs
Cell 196
(GrpE/DnaJ) have yielded the most comprehensive view to date of the roles of HSPs in protein assembly, and provide another example of cooperation among HSPs (reviewed by Georgopoulos et al., 1990). Steps in h DNA replication include the following: the h0 initiator protein binds specifically to or&, the hP initiator protein complexes with the E. coli DnaB helicase, and the hP-DnaB complex becomes localized to orih through an ho-hP interaction. The overall role of HSPs in DNA replication is the dissociation of the protein complex located at orih and the sequestration of the hP protein, allowing activation of DnaB and the initiation of replication. Zylicz (University of Gdansk) showed that DnaJ and GrpE stimulate the ATPase activity of DnaK, modulate conformational changes in DnaK triggered by ATP hydrolysis, and modulate the stability of the hP-DnaK complex. DnaJ stimulates the rate of ATP hydrolysis by DnaK, and GrpE increases the rate of ADP release from DnaK (K. Liberek, University of Utah). Additional evidence was presented that a DnaJ-DnaK-ATP interaction is needed to release protein P from preinitiation complexes (Hoffmann, University of California, Berkeley). Using a phage Pi-based model of DNA replication, Wickner (NIH) showed that DnaK and DnaJ activate the RepA replication initiation protein by converting RepA dimers to monomers. This unexpected finding is particularly exciting because it suggests possible analogies with the c-Jun and c-Fos transcription factors, whose activities appear to be modulated by conversion of homodimers to more potent heterodimers. The DnaWHsp70 proteins may facilitate such conversions, allowing cells to respond quickly to new environmental signals. Acquired Thermotolerance and Thermoresistance When either embryos or cultured cells are subjected to a mild heat stress or to certain chemical stressors, they acquire the capacity to survive what would normally be lethal exposures (reviewed by Hahn and Li, 1990; Petersen, 1990). Virtually every major class of HSPs has now been implicated in this acquired thermotolerance. Lindquist (University of Chicago) showed that Hsp70 plays a major role in thermotolerance in Drosophila based on mutational analyses, overexpression of Hsp70, and inhibition of Hsp70 expression using antisense RNA. She also reported on studies of yeast deletion mutants indicating that Hspl04 plays a major role in induction of thermotolerance. Interestingly, the mutants still retain a transient capacity to tolerate heat, an indication of the multifaceted nature of this response. Li (Sloan-Kettering Cancer Center) used a murine retroviral vector to express constitutively a human hsp70 gene in rodent cells. Stable transformants displayed increased levels of resistance to thermal killing that correlated with the amounts of Hsp70 present. In other studies, thermotolerant mammalian cells accumulated less protein damage than unprotected cells, and Hsp70 was implicated in protection/repair of protein damage in thermotolerant cells (Kim, Stanford University). Evidence was presented that bovine Hsc70, added exogenously to cultured arterial cells and injected into rat retina, may prolong arterial cell survival in culture and protect retinal cells from light damage (Tytell, Wake Forest University).
Induction of Hsp27 alone, using an inducible expression vector driven by a metallothionein promoter, rapidly conferred protection against thermal killing on mouse cells in culture (Lavoie, University of Laval). Phosphorylation of preexisting Hsp27 is an early event in mammalian cellular responses to stress and was suggested as a possible signal for development of thermotolerance (Landry, University of Laval). The a-crystallins of vertebrate lenses belong to the same superfamily as small HSPs like Hsp27. It has now been shown that aB-crystallin is an HSP and the promoter of its gene contains a heat shock element. Remarkably, overexpression of aB-crystallin in mammalian cells leads to the acquisition of thermotolerance (Chepelinsky, NIH; Klemenz, University of Zurich). Then, there are those disturbing instances when HSPs and thermotolerance come apart. For example, yeast cells with a mutation in their HSF gene are defective for growth and mitochondrial protein import at elevated temperature, but are still able to acquire thermotolerance (Smith, University of California, San Diego). A mouse lymphoma B cell line acquires thermotolerance in culture and when grown as an ascites tumor, but manages to induce HSPs only in the latter case (Anderson, Peter MacCallum Cancer Institute). Conversely, induction of HSPs in E. coli without a temperature shift is insufficient to induce thermotolerante (VanBogelen, University of Michigan). Obviously, understanding thermotolerance remains for the future. Fishing for Genetic Variation in Streams and Culture Dishes Another future challenge will be to apply our knowledge of stress responses to the natural environment. For example, the impact of the Big Heat Shock, i.e., rapid global warming, and other forms of pollution may depend on how much genetic variation in resistance mechanisms is available in natural populations. This is largely unknown. Clonal reproduction among certain hybrid species of the Sonoran Desert topminnow Poeciliopsis allows amplification and study of naturally occurring combinations of genetic alleles in the laboratory. This has led to evaluation of the deploy ment of genetic resistance mechanisms such as the heat shock response in natural populations (Schultz, 1989). Hightower showed that these desert fish have a highly exaggerated heat shock response, presumably part of their adaptation, and different species have elaborated different isoforms of the Hsp70 and Hsp20-30 families that may contribute to observed differences in thermal resistance among biotypes. Several groups described what may be cell culture versions of the experiments of nature described above. Anderson and Hahn (Stanford University) selected heatresistant variants of mammalian cells that express elevated levels of the major HSPs and have a new Hsp70 family isoform. Laszlo (Washington University) described heat-sensitive variants unable to develop thermotolerance and unable to recover rapidly from thermal damage, and heat-resistant variants that overexpressed Hsp70 and Hsp27. These data suggest that thermal adaptation at the cellular level is associated with changes in the heat shock response. Gupta (McMaster University) showed that mu-
Meeting 197
Review
tants of a mammalian cell line selected for resistance to microtubule inhibitors contain new isoforms of Hsp60 and Hsc70, suggesting that they may interact with tubulins during microtubule assembly (Gupta, 1990). These studies taken together raise the possibility that adaptation to stress by cells in culture and organisms in the environment may require changes in chaperones. The exploration of these changes may yield new insights into the evolution of protein folding and assembly pathways. Acknowledgments I wish to thank Costa Georgopoulos, David McKay, Rick Morimoto, Seth Sadis, and Bill Welch for reading and commenting on the manuscript. Unfortunately, there were many more interesting presentations than could be fit into this review. References Beckmann, Ft. P., Mizzen, L. A., and Welch, W. J. (1990). Interaction of Hsp70 with newly synthesized proteins: implications for protein folding and assembly. Science 248, 850-854. Craig, E. A., and Gross, C. A. (1991). Is hsp70 the cellular ter? Trends Biochem. Sci. 76, 135-139.
thermome-
Creighton, T. E. (1988). Toward a better understanding of protein ing pathways. Proc. Natl. Acad. Sci. USA 85, 5082-5086.
fold-
Edington, B. V., Whelan, S. A., and Hightower, L. E. (1989). Inhibition of heat shock (stress) protein induction by deuterium oxide and glycerol: additional support for the abnormal protein hypothesis of induction. J. Cell. Physiol. 739, 219-228. Ellis, Ft. J. (1990).
Molecular
chaperones.
Semin.
Cell Biol. 7, I-72.
Flaherty, K. M., DeLuca-Flaherty, C., and McKay, D. B. (1990). Threedimensional structure of the ATPase fragment of a 70K heat-shock cognate protein. Nature 346, 623-628.
of bacteria by differential scanning denaturation in situ to maximum phys. Acta 7055, 19-26.
calorimetry: relationship of protein growth temperature. Biochim. Bio-
Lupas, A., Van Dyke, M., and Stock, J. (1991). from protein sequences. Science 252, 1162-l
Predicting 164.
coiled coils
Mizzen, L. A., Kabiling, A. N., and Welch, W. J. (1991). The two mammamlian mitochondrial stress proteins, grp75 and hsp58, transiently interact with newly synthesized mitochondrial proteins. Cell Reg. 2, 165-179. 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). Neupert, W., Hartl, F.-U., Craig, E. A., and Pfanner, polypeptides cross the mitochondrial membranes? Nover, L. (1991). CRC Press).
The
Heat Shock
Response
(Boca
N. (1990). How do Cell 63, 447-450. Raton,
Florida:
Parsell, D. A., and Sauer, R. T. (1989). Induction of a heat shocklike response by unfolded protein in Eschericbia co/i: dependence on protein level not protein degradation. Genes Dev. 3, 1226-1232. Pelham, H. R. B. (1990). Functions of the hsp70 protein family: an overview. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 287-299. Petersen, N. S. (1990). Effects of heat and chemical stress on development. In Advances in Genetics: Genomic Responses to Environmental Stress, J. G. Scandalios, ed. (New York: Alan R. Liss), pp. 275-296. Sadis, S., Raghavendra, K., and Hightower, L. E. (1990). Secondary structure of the mammalian 70-kilodalton heat shock cognate protein analyzed by circular dichroism spectroscopy and secondary structure prediction. Biochemistry 29, 8199-8206. Schultz, R. J. (1989). Origins and relationships of unisexual poeciliids. In Evolution and Ecology of Poeciliid Fishes, G. K. Meffe and F. F. Snelson, Jr., eds. (Englewood Cliffs, New Jersey: Prentice-Hall), pp. 69-87.
Fletterick, Ft. J., Bates, D. J., and Steitz, T. A. (1975). The structure of a yeast hexokinase monomer and its complexes with substrates at 2.7 A resolution. Proc. Natl. Acad. Sci. USA 72, 38-42.
Skowyra, D., Georgopoulos, C., and Zylicz, M. (1990). The E. coli dnaK gene product, the hsp70 homolog, can reactivate heat-inactivated RNA polymerase in an ATP hydrolysis-dependent manner. Cell 62, 939-944.
Gaitanaris, G. A., Papavassiliou, A. G., Rubock, P., Silverstein, S. J., and Gottesman, M. E. (1990). Renaturation of denatured h repressor requires heat shock proteins. Cell 67, 1013-1020.
Sorger, P. K. (1991). Cell 65, 363-366.
Georgopoulus, C., Ang, D., Liberek, K., and Zylicz, M. (1990). Properties of the Escherichia co/i HSPs and their role in bacteriophage I growth. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 191-221. Goloubinoff, P., Christeller, J. T., and Gatenby, A. A. (1989). Reconstitution of active dimeric ribulose bisphosphate carboxylase from an unfolded state depends on two chaperonin proteins and Mg-ATP. Nature 342, 884-889. Gupta, R. S. (1990). the in viva assembly 418.
Mitochondria, molecular chaperone of microtubules. Trends Biochem.
proteins and Sci. 75, 415-
Hahn, G. M., and Li, G. C. (1990). Thermotolerance, thermoresistance, and thermosensitization. In Stress Proteins in Biology and Medicine, R. I. Morimoto, A. Tissieres, and C. Georgopoulos, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 79-100. Kabsch, W., Mannherz, (1990). Atomic structure 37-44.
H. G., Suck, D., Pai, E. F., and Holmes, K. C. of the actin:DNase I complex. Nature 347,
Kozutsumi, Y., Segal, M., Normington, K., Gething, M.J., and Sambrook, J. (1988). The presence of malfolded proteins in the endoplasmic reticulum signals the induction of glucose-regulated proteins. Nature 332, 462-464. Lepock, J. R., Frey, H. E., Rodahl, A. M., and Kruuv, J. (1988). Thermal analysis of CHL V79 cells using differential scanning calorimetry: implications for hyperthermic cell killing and the heat shock response. J. Cell. Physiol. 737, 14-24. Lepock,
J. R., Frey, H. E., and Inniss,
W. E. (1990). Thermal
analysis
Heat shock
factor
and the heat shock
response.
VanBogelen, R. A., and Neidhardt, F. C. (1990). Ribosomes as sensors of heat and cold shock in Escherichia co/i. Proc. Natl. Acad. Sci. USA 87. 5589-5593. Note
Added
Induction changes
in Proof
of HSPs can also be modulated in Drosophila in glutamine concentrations (Sanders, UMDNJ).
cells
by
