Hsp90's secrets unfold: new insights from structural and functional studies

reviews 35 Mosquera, L. et al. (1993) Development 117, 377–386 36 Nagase, T. et al. (1996) DNA Res. 3, 321–329 37 Stebbins-Boaz, B., Hake, L. E. and Richter, J. D. (1996) EMBO J. 15, 2582–2592 38 Zhao, G. Q., Liaw, L. and Hogan B. L. (1998) Development 125, 1103–1112 39 Zhao, G. Q. and Hogan, B. L. (1997) Mech. Dev. 61, 63–73 40 Lilly, M. A. and Spradling, A. C. (1996) Genes Dev. 10, 2514–2526 41 Hawkins, N. C., Thorpe, J. and Schupbach, T. (1996) Development 122, 281–290 42 Gönczy, P., Matunis, E. and DiNardo, S. (1997) Development 124, 4361–4371 43 Matunis, E. et al. (1997) Development 124, 4383–4391 Hsp90’s secrets unfold: new insights from structural and functional studies Avrom J. Caplan Hsp90 is a molecular chaperone associated with the folding of signal-transducing proteins, such as steroid hormone receptors and protein kinases. Results from recent studies have shed light on the structure of Hsp90 and have demonstrated that it can bind to and hydrolyse ATP. Hsp90 forms several discrete subcomplexes, each containing distinct groups of co-chaperones that function in folding pathways. Although Hsp90 is not generally involved in the folding of nascent polypeptide chains, there is a growing list of proteins whose activity depends on its function, including heat-shock factor. This review addresses recent developments in our understanding of the structure and function of Hsp90. Molecular chaperones are now accepted as being a general feature of the cellular protein-folding landscape. They integrate translation with subsequent steps in the life cycle of polypeptide chains, such as folding in the cytosol or translocation across endoplasmic reticulum and mitochondrial membranes. Several major classes of molecular chaperones have been identified and characterized as polypeptide- 262 44 Braun, R. E. et al. (1989) Nature 337, 373–376 45 Pauli, D. and Mahowald, A. P. (1990) Trends Genet. 6, 259–264 46 Swain A. et al. (1998) Nature 391, 761–767 47 Jenuth, J. P. et al. (1996) Nat. Genet. 14, 146–151 48 Yu, R. N. et al. (1998) Nat. Genet. 20, 353–357 49 Lyon, M. F. et al. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 6968–6972 50 Lehmann, A. L. et al. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 9436–9441 51 Hardy, R. W. et al. (1979) J. Ultrastruct. Res. 69, 180–190 52 Giardinia, A. (1901) Int. Mschr. Anat. Physiol. 18, 417–484 53 Jörgensen, M. (1913) Arch. Zellforsch. 10, 1–126 binding proteins with low sequence specificity. They bind to stretches of polypeptide rich in hydrophobic amino acids such as those found in unfolded proteins. They function by repeated cycles of polypeptide binding and release, and facilitate folding by preventing or reversing off-pathway interactions. Some molecular chaperones, such as the chaperonins, provide a sequestered environment so that a polypeptide can fold unperturbed by other events in the cytosol or mitochondrial matrix. Molecular chaperones do not alter the folding fate of a polypeptide – rather, they shift the equilibrium of the folding reaction towards the folded state and away from the aggregated state1,2. Hsp70 and the prokaryotic/organelle chaperonins fit this definition of a molecular chaperone quite well: both are abundant, known to be required in protein-folding reactions and are induced by heat stress – a condition that causes protein unfolding1,2. Hsp90, on the other hand, is also an abundant molecular chaperone that is highly induced by heat stress, yet, compared with Hsp70, is not well understood. For example, Hsp90 has been the focus of many studies over the past 15 years or so, primarily as a chaperone involved in the maturation of signal-transducing proteins, notably steroid hormone receptors and protein kinases3. What has been unclear, however, is whether Hsp90 has a broader role in polypeptide dynamics within the cell and whether it functions in a similar manner to other known molecular chaperones. For example, there was a long-standing controversy about whether Hsp90 could bind to and hydrolyse ATP, as do other molecular chaperones such as Hsp70 and the chaperonins. However, results from recent studies have shed new light on the subject and have clarified many areas of uncertainty. Hsp90 structure Hsp90 is a cytosolic protein that is one of a small family of molecular chaperones. Prokaryotic orthologues have been characterized, and it is also abundant in the endoplasmic reticulum of eukaryotes (where it is known as grp94). Hsp90 is a dimeric protein (monomers are 80–90 kDa, depending on the source) with an elongated structure. Recent studies4 on grp94 revealed that is has a trinodular appearance, 0962-8924/99/$ – see front matter © 1999 Elsevier Science. All rights reserved. PII: S0962-8924(99)01580-9 trends in CELL BIOLOGY (Vol. 9) July 1999 reviews with dimensions 28 nm by 7 nm. Since the dimerization domain is at the C-terminus4,5, this arrangement is thought to reflect an end-to-end arrangement, leaving the N-termini opposed (Fig. 1). A similar arrangement was described recently for cytosolic Hsp905, which is also capable of undergoing dramatic structural changes at high temperature, or in the presence of ATP, to form a closedcircular structure. The 25-kDa N-terminal domain has been crystallized independently by two groups and found to contain binding sites for both nucleotide and geldanamycin, an anti-tumour drug that inhibits Hsp90 action6–8. The structure of the N-terminal domain is an a–b sandwich with a ligand-binding pocket (Fig. 2). On the ‘floor’ of this structure are eight uninterrupted antiparallel b sheets. The pocket sides are formed by FIGURE 1 nine a helices and intervening loops. Hsp90 structure. (a) Different domains of Hsp90 labelled according to function (where known). Binding of nucleotide to Hsp90 differs (b) Montage of images of the endoplasmic reticulum Hsp90 – grp94 – as viewed in the electron somewhat compared with Hsp70. First, microscope (note the trinodular appearance; images courtesy of Chris Nicchitta). Bar, 25 nm. the binding of nucleotide to Hsp90 is (c) Schematic representation of Hsp90 and its possible range of movements from a linear molecule (i), much weaker than for Hsp70: the dis- through a bent intermediate [ii; see fourth image in (b)] into a circular structure (iii; see Ref. 5). sociation constant of ADP for Hsp70 is with the circular structures recently observed by 0.1 mM (Ref. 9) and for Hsp90 it is 29 mM (Ref. 8.) The Yahara and colleagues5. Interestingly, the N-termidissociation constant for ATP is also submicromolar nal dimers have a large groove provided by a confor Hsp70, whereas it is over 100 mM for Hsp90. Both tinuum of b sheets from each monomer (see Fig. 2). proteins display very slow rates of ATP hydrolysis in The groove width is large enough to accept an unthe absence of co-chaperones. Furthermore, the folded peptide, although whether such a structure is topology of nucleotide bound to Hsp90 is quite unphysiologically relevant is still an open question. like that for Hsp70, specifically with respect to the accessibility of the C-8 position in the adenine ring. Role of Hsp90 in protein folding Experiments to detect nucleotide interactions with Several recent investigations have focused on Hsp70 have commonly used derivatives that are whether Hsp90 is important for the folding of conjugated via a C-8 linkage, but this position is nascent polypeptide chains and/or for polypeptide sterically hindered in ATP bound to Hsp908. This refolding after heat stress. However, loss of Hsp90 difference probably accounts for difficulties in defunction in yeast, using a conditional temperaturetecting nucleotide interactions with Hsp9010. sensitive mutant, did not affect protein folding or Whether ATP binding and hydrolysis are essential refolding in a general manner, although modest for Hsp90 function in vivo was addressed recently defects were observed in the folding of some heterusing the yeast system11,12. For these experiments, ologously expressed proteins13. Similarly, studies amino acids crucial to nucleotide binding (Asp79) performed in vitro with the test substrates luciferase and hydrolysis (His33) were deduced from the crysand b-galactosidase failed to find an essential role tal structure and then mutated. Genes encoding for Hsp90 in the folding reaction, although it was these mutant versions of Hsp90 were introduced capable of stimulating folding by other molecular into yeast strains that were deleted for endogenous chaperones, such as Hsp70 and Hsp40 (dnaJ)14,15. Hsp90 genes but kept viable by a plasmid-borne verThus, Hsp90 is not essential for protein folding in a sion of the wild-type gene. Upon selection of yeast general sense, although these studies do not rule out strains that lose the plasmid-borne wild-type gene, the possibility that Hsp90 might have a wider range of the ability of the mutants to rescue the lethal substrates than is currently recognized. Furthermore, Avrom J. Caplan is phenotype of an Hsp90-null strain could be deterit also remains unclear how much functional redun- in the Dept of Cell mined. However, since neither mutant rescued dancy exists between Hsp90 and other chaperone Biology and the lethal phenotype, it was concluded that Hsp90 systems in protein folding. requires its ATPase to function properly. Anatomy, Mount Several studies have monitored protein activity Sinai School of An intriguing observation made by Pearl and colafter treatment with the Hsp90 inhibitor gel- Medicine, New leagues was that the N-terminal nucleotide-binding danamycin. This reagent (or similar compounds of York, NY 10029, domain of Hsp90 was a dimer in their crystals6,8. the same class of ansamycin antibiotics) blocks the USA. This orientation is at odds with the biochemical folding of protein kinases and steroid hormone E-mail: acaplan@ characterization showing that Hsp90 dimerized via receptors in vitro16,17 and stimulates degradation of smtplink.mssm.edu its C-terminal domains, although it is consistent trends in CELL BIOLOGY (Vol. 9) July 1999 263 reviews FIGURE 2 Crystal structure of the 25-kDa N-terminal domain of Hsp90, shown with the ligands geldanamycin (GA) and ATP. Two orientations of each structure are shown in the top panel, the side view in the upper structure being turned 908 towards the reader to reveal a top view in the lower structure. a Helices are shown in yellow, b sheets in blue and ligand in red. The GAbound structure is human Hsp90, whereas the ATP-bound structure is yeast Hsp90. The lower panel shows the structure of an N-terminal dimer of yeast Hsp90 as described in Refs 6 and 8. The a-carbon backbone is shown except for the b sheets (blue). several different polypeptides, including protein kinases18,19, steroid receptors20,21, membrane proteins [such as apolipoprotein B22, cystic fibrosis transmembrane conductance regulator (CFTR)23 and several receptor tyrosine kinases24,25], p5326 and luciferase27 in vivo. This degradation is mediated at least partially by the proteasome, and there is some evidence for physical association between Hsp90 and this multicatalytic protease28. However, geldanamycin also stimulates expression of Hsp7029, suggesting that there might not be a direct relationship between Hsp90 loss of function and protein degradation. Indeed, Hsp70 overexpression alone can also stimulate protein degradation, suggesting that the effect of geldanamycin might be indirect22. The ability of Hsp90 to stimulate folding by other chaperones suggests that it can bind to peptides by itself, and it appears to be capable of stabilizing unfolded proteins and protecting them from aggregating. To do this, Hsp90 would need to have its own peptide- or polypeptide-binding sites. These have been shown to exist in two domains of Hsp90, at the N-terminus and at the C-terminus30,31, with each 264 having a different substrate preference. Binding of peptides to the N-terminal domain is also destabilized by ATP and geldanamycin, whereas peptide binding to the C-terminal domain is unaffected by either reagent. Peptide binding by grp94 might also have an important clinical application for the treatment of cancer. Srivastava and colleagues demonstrated recently that grp94 purified from specific tumours could be used for subsequent immunotherapy when targeted against the same specific tumour, but not other tumour types32. This specificity lies in the unique complement of peptides that copurify with grp94 when prepared from different sources. Peptide binding by grp94 might reflect a normal cellular function related to maturation of major histocompatibility complexes within the endoplasmic reticulum. Recent studies have shown that peptide binding by grp94 occurs in a hydrophobic environment of the protein and causes conformational shifts in its tertiary structure33. Thus, Hsp90 appears to behave as a molecular chaperone: it can bind to peptides, prevent protein aggregation and can function, in a limited capacity, in protein-folding reactions. These processes probably occur via conformational changes in Hsp90 that are driven by nucleotide binding and hydrolysis34, and, although it is unlikely to have a general role in cellular protein folding, there is abundant evidence to show that it is important for the folding of protein kinases and steroid hormone receptors. Indeed, two lessons learned from studies on steroid hormone receptor maturation have provided a solid background for understanding the role of Hsp90 in protein folding. The first is that Hsp90 does not act alone but in association with several different ‘cochaperones’. They bind to Hsp90 and organize it into discrete subcomplexes, each containing a different set of co-chaperones. The second lesson is that the co-chaperones appear to function along a folding pathway that has a defined order35. Hsp90 and its co-chaperones A list of different Hsp90 co-chaperones is given in Table 1, and their organization into discrete subcomplexes is shown in Fig. 3. To date, four different subcomplexes have been described, two of which are associated with maturation of steroid hormone receptors; these are the Hsp90–Hop(Sti1)–Hsp70–Hip complex (which I will refer to as the ‘Hop’ complex) and the Hsp90–p23/immunophilin complex (which I will refer to as the ‘p23’ complex). The Hop complex functions upstream of the p23 complex, although the exact relationship between all of the proteins in these subcomplexes is unclear. However, what is known is the minimal complement of proteins necessary to refold the hormone-binding domains of the progesterone and glucocorticoid receptors efficiently; this is Hsp90, Hop, Hsp70, p23 and a dnaJ protein36,37. The dnaJ proteins are regulators and co-chaperones of Hsp7038, and their function in steroid receptor and v-Src signalling was first described in studies using yeast as a model system39–41. They are not stable components of Hsp90 chaperone complexes and appear to function catalytically on trends in CELL BIOLOGY (Vol. 9) July 1999 reviews TABLE 1 – HSP90 AND HSP70 CHAPERONES AND CO-CHAPERONES Higher eukaryote Yeasta Hsp90 Hsp70 Hsp82p/Hsc82p Ssa1p–Ssa4p Hop Hip Bag1 Sti1p N/A N/A Hsp90 Hsp70 TPR co-chaperone co-chaperone motifs No No ✔ Hsp40/Hdj2/Hsj1d Ydj1p/Sis1pd Immunophilinsb p23 complexes p50 ?c Cpr6p/Cpr7p Sba1p ✔ ✔ Cdc37p Cns1p ✔ ✔ Yes No Molecular chaperone Molecular chaperone, broad peptide binding specificity Mediates Hsp90–Hsp70 interaction Stabilizes Hsp70 in ADP-bound state Nucleotide-exchange factor for Hsp70 Catalyses ATP hydrolysis in Hsp70. Targets Hsp70 to substrates Peptidyl-prolyl isomerases Stabilizes Hsp90 substrate No Yes Molecular chaperone Binds Hsp90 in complex with Hsp70 ✔ ✔ ✔ Yes Yes No ✔ No ✔? Function a N/A: not known in yeast. There are several different mammalian immunophilins that bind to Hsp90; see Ref. 3 for details. c Cns1p was discovered recently, and it is unknown whether vertebrate homologues exist. d Eukaryotic dnaJ proteins. b Hsp70 in the context of steroid receptor folding. It is thought that their function is to target Hsp70, which then subsequently targets Hsp90 (via Hop) to conformationally immature steroid receptor molecules36,37. The proposed Hop requirement is based on the inability of Hsp70 to bind directly to Hsp90 unless Hop is present42. A model of these events is shown in Fig. 4. A eukaryotic dnaJ protein stimulates stable association of Hsp70 with a steroid receptor by catalysing its ATPase (steps 1 and 2). The receptor–Hsp70 complex then interacts with Hsp90 via Hop, which binds to both proteins and maintains Hsp90 in its ADP-bound form43,44 (step 3). The receptor–Hop complex is unstable, however, and rapidly dissociates unless p23 is present (step 4). However, p23 can only bind to Hsp90 in the presence of ATP34, contrasting with the binding conditions for Hop and suggesting that nucleotide exchange in Hsp90 must take place in association with this event. Thus, p23 and Hop reside on separate complexes because their binding is stabilized by different nucleotide-binding states of Hsp90. Several other co-chaperones such as Hip (a component of Hsp70 in the Hop complex) and immunophilins (components of the p23 complex) are present in vivo but are not essential components of the pathway in mammalian systems (see Ref. 3 for a comprehensive review). A yeast cyclophilin (Cpr7p) has been described as being important for the maturation of glucocorticoid receptors expressed in yeast45. Although Hop and p23 reside on Hsp90 molecules with different nucleotide states, their relationship in the context of steroid receptor folding is likely to be more dynamic. It seems unlikely that entire subcomplexes would replace each other in the course of steroid receptor maturation; instead, key individual co-chaperones could move dynamically into the complex and then serve to recruit those trends in CELL BIOLOGY (Vol. 9) July 1999 that define each subcomplex at steady state. Thus, although the majority of p23 and Hop reside on separate Hsp90 molecules, there are situations in which p23 can associate with Hsp90 bound to Hop37,46,47. Pratt and colleagues dubbed such a dynamic chaperone–‘substrate’ complex the ‘foldosome’ (see Ref. 46). The importance of p23 in steroid hormone receptor maturation suggests that it might have a general role in Hsp90 function. Recent results from studies in yeast are not consistent with this view, however, since deletion of the Saccharomyces cerevisiae p23 homologue, SBA1, had only a very mild effect on growth or folding of other known targets of Hsp90, including steroid hormone receptors and v-Src47,48. However, overexpression of SBA1 did have the remarkable effect of increasing the activity of heterologously expressed steroid hormone receptors without changing protein levels (B. C. Freeman, pers. commun.; Ref. 49). This probably reflects the existence of two pools of receptor when expressed in yeast – one active and the other inactive50. The equilibrium between the two might be dependent on Hsp90 and its co-chaperones, in this case p23/Sba1p. If such co-chaperones are limiting, then their overexpression could shift the equilibrium from the inactive state (presumably aggregated or otherwise misfolded) to the active state. The binding site on Hsp90 for several of the co-chaperones resides within its C-terminal dimerization domain51,52. The co-chaperones that bind in this region are all related by having tetratricopeptide repeat (TPR) sequences, which mediate their association, in a competitive manner, with Hsp9053. However, it seems unlikely that such competition exclusively defines different Hsp90-containing subcomplexes because these might also depend on the nucleotide state of Hsp90 and the presence of nonTPR-containing co-chaperones. Furthermore, at least one Hsp90-containing subcomplex in yeast binds to two different TPR co-chaperones simultaneously – 265 reviews FIGURE 3 Hsp90 co-chaperones and complexes. Four complexes are shown. (a) The Hop complex contains Hop, which binds to both Hsp90 and Hsp70, and Hip, a co-chaperone of Hsp70. Another Hsp70 co-chaperone, called Bag1, might also be present in these complexes but binds competitively with Hip66. Note that Hop is shown interacting with both C-terminal and N-terminal domains of Hsp90 as described in Refs 44, 51 and 52. (b) The p23 complex comprises p23 (or Sba1p in yeast) and one of several immunophilins (Cyp40 is shown; see Ref. 3 for more details). p23 binds to Hsp90–ATP, but its exact binding site is unknown. (c) Cdc37 binds to Hsp90 in the absence of other known co-chaperones. It is unclear how the nucleotide state of Hsp90 affects Cdc37 binding. (d) The CNS1 complex is unique in that it contains two different tetratricopeptide repeat (TPR)-containing cochaperones, Cns1 and Cpr7. Hsp70 is also a component of this complex. Cpr7p (an immunophilin) and Cns1p (which is related to Sti1p/Hop; see below) – in a complex that also contains Hsp7054,55. There is also evidence that yeast Hop (Sti1p) binds to Hsp90 as a dimer, suggesting that there are two binding sites for TPR domains in the native protein44. This could create the asymmetry required for binding of two different TPR co-chaperones, provided they do not bind in a dimeric form. The characterization of Cns1p also solved a mystery concerning the association of Hsp70 with Hsp90 in yeast. As mentioned above, Hsp90 and Hsp70 interact indirectly via Hop. Deletion of the yeast Hop homologue (called Sti1p) did not affect the levels of Hsp70 bound to Hsp9056, suggesting the existence of other factors that fulfil this role. Thus, Hsp70 is a component of the Hop complex and also of a complex containing Cns1p. Higher-eukaryotic homologues of Cns1p have yet to be described. The TPR motif of some co-chaperones is required for binding to Hsp90, but it might also serve another function. Yeast has two TPR-containing cyclophilins, Cpr6p and Cpr7p, that bind to Hsp90 in complexes with Sba1p (yeast p23) and Cns1p, respectively47,54,55. These two Hsp90-bound yeast cyclophilins are structurally similar but functionally distinct. Deletion of 266 CPR6 produces no observable phenotype, whereas deletion of CPR7 results in slow growth. In addition, overexpression of CPR6 in a cpr7D strain did not suppress the growth phenotype54,55. Studies addressing the functional domains of Cpr7p found, surprisingly, that the TPR repeats themselves can substitute for the whole Cpr7p protein, rather than the peptidylprolyl isomerase domain57. The nature of this other function remains to be determined. The binding sites for non-TPR-containing cochaperones are understood poorly. Binding by yeast or vertebrate p23 is affected by multiple mutations across the length of Hsp9034,47. Association of Cdc37p with Hsp90 has also been examined, and its binding site appears to overlap, although being distinct from, the binding site for TPR proteins58. Cdc37p also functions independently of Hsp9058,59, and it remains unclear whether it forms stable direct complexes with yeast Hsp90. Cdc37p functions in the folding of protein kinases and is important for some cyclin–CDK (cyclin-dependent kinase) interactions60. Several studies suggest that it does not affect the folding of the steroid hormone receptors, although it appears to be important for hormone-dependent activation of the androgen receptor when it is expressed in yeast but not of the glucocorticoid receptor61. Hsp90 and heat-shock factor The hypothesis that molecular chaperones regulate their own induction has been proposed previously, with Hsp70 being the most likely chaperone to fulfil that role. However, recent studies suggest that Hsp90 is an important negative regulator of the heat-shock response in both lower and higher eukaryotes. Interactions between Hsp90 and the heat-shock transcription factor (HSF) have been reported, but two new studies have now characterized the functional relevance of this association by using a combination of genetic and biochemical methodologies. Using extracts from human cells and a DNA mobility-shift assay system, Zou et al.29 showed that HSF binding to DNA was stimulated by reagents that inhibit or reduce the endogenous levels of Hsp90. Similar results were observed in yeast58, where it was shown that depletion or mutation of Hsp90 resulted in increased synthesis of proteins under direct transcriptional control of HSF62. In addition, it was also found that deletion of the yeast cyclophilin 40 homologue CPR7 also leads to HSF activation62, although depletion of the human cyclophilin 40 homologue had no effect in the experiments reported by Zou et al.29 These results explain the findings of several investigators who observed that geldanamycin treatment of cells increased the synthesis of Hsp70. Hsp90 and evolution The heat-shock field was initiated by studies on Drosophila, and a new investigation in the same organism has finally produced some idea as to how loss of Hsp90 function affects a multicellular eukaryote. In this new work63, Rutherford and Lindquist observed multiple phenotypic variations in flies containing mutant Hsp90, or Hsp90 inhibited by geldanamycin. They proposed that Hsp90 trends in CELL BIOLOGY (Vol. 9) July 1999 reviews normally suppresses genetic variation by acting to buffer its expression, possibly at the level of protein folding. However, once the phenotypic variation occurs, it becomes independent of Hsp90 function, and even overexpression of Hsp90 fails to revert the phenotype. Rutherford and Lindquist argued that evolutionary change itself could be stimulated by conditions that might transiently cause a decrease in the cellular level of Hsp90. Such conditions might occur during heat stress, as Hsp90 becomes associated with heat-damaged proteins (and before heatinducible expression takes place). This hypothesis was supported by the finding that the frequency of phenotypic variation in the Hsp90 mutant strains was increased under stressful conditions. Concluding remarks The recent studies discussed above do not provide a unifying hypothesis for how Hsp90 functions. Rather, they show how variable it is in terms of both structure and function as a chaperone machine. The major unanswered questions regarding Hsp90 are similar to those being asked of other chaperone systems: how wide is the range of proteins that require Hsp90 for their activity, and what roles do Hsp90 cochaperones play in the folding reaction? The original and rather narrowly defined family of substrates – steroid hormone receptors and protein kinases – is slowly being expanded to include new members, such as nitric oxide synthase64 and telomerase65. New research will undoubtedly concentrate on the dynamic motions of Hsp90 and their relationship to the folding reaction. The roles of Hsp90 co-chaperones are also being revealed at great speed, and the finding of novel Hsp90 co-chaperones, such as yeast Cns1p, suggests that Hsp90 is much more complex than had been imagined previously. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Bukau, B. and Horwich, A. L. (1998) Cell 92, 351–366 Hartl, F. U. (1996) Nature 381, 571–580 Pratt, W. B. and Toft, D. O. (1997) Endocr. Rev. 18, 306–360 Wearsch, P. A. and Nicchitta, C. V. (1996) Biochemistry 35, 16760–16769 Maruya, M. et al. (1999) J. Mol. Biol. 285, 903–907 Prodromou, C. et al. (1997) Nat. Struct. Biol. 4, 477–482 Stebbins, C. E. et al. (1997) Cell 89, 239–250 Prodromou, C. et al. (1997) Cell 90, 65–75 McKay, D. B. et al. (1994) in The Biology of Heat Shock Proteins and Molecular Chaperones (Morimoto, R. I., Tissieres, A. and Georgopoulos, C., eds), pp. 153–177, Cold Spring Harbor Press Jakob, U. et al. (1996) J. Biol. Chem. 271, 10035–10041 Panaretou, B. et al. (1998) EMBO J. 17, 4829–4836 Obermann, W. M. J. et al. (1998) J. Cell Biol. 143, 901–910 Nathan, D. F., Vos, M. H. and Lindquist, S. (1997) Proc. Natl. Acad. Sci. U. S. A. 94, 12949–12956 Schumacher, R. J. et al. (1996) Biochemistry 35, 14889–14898 Freeman, B. C. and Morimoto, R. (1996) EMBO J. 15, 2969–2979 Hartson, S. D. et al. (1996) Biochemistry 35, 13451–13459 Smith, D. F. et al. (1995) Mol. Cell. Biol. 15, 6804–6812 Whitesell, L. et al. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 8324–8328 Schulte, T. W. et al. (1995) J. Biol. Chem. 270, 24585–24588 Whitesell, L. and Cook, P. (1996) Mol. Endocrinol. 10, 705–712 trends in CELL BIOLOGY (Vol. 9) July 1999 FIGURE 4 A model for Hsp90-mediated folding of a steroid hormone receptor ligand-binding domain. The receptor first interacts with Hsp70 and a eukaryotic dnaJ protein (J; step 1). Hsp70 binding to the receptor is stabilized by dnaJ-stimulated hydrolysis of ATP (step 2). Hsp70–ADP then binds to Hop, which is also bound to Hsp90, thus bringing Hsp90 into direct contact with the receptor (step 3). This complex is unstable, however, and these co-chaperones are replaced by p23 and nucleotide exchange in Hsp90 (step 4). The receptor can then cycle (step 5) to a state that is not complexed with molecular chaperones. Subsequent unfolding and/or aggregation of the unliganded receptor then leads to molecular chaperone binding. (Model adapted from Ref. 37.) 21 Segnitz, B. and Gehring, U. (1997) J. Biol. Chem. 272, 18694–18701 22 Fisher, E. A. (1997) J. Biol. Chem. 272, 20427–20434 23 Loo, M. A. et al. (1998) EMBO J. 17, 6879–6887 24 Sepp-Lorenzino, L. et al. (1995) J. Biol. Chem. 270, 16580–16587 25 Imamura, T. et al. (1998) J. Biol. Chem. 273, 11183–11188 26 Whitesell, L. et al. (1997) Oncogene 14, 2809–2816 27 Schneider, C. et al. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 14536–14541 28 Wagner, B. J. and Margolis, J. W. (1995) Arch. Biochem. Biophys. 323, 455–462 29 Zou, J. et al. (1998) Cell 94, 471–480 30 Young, J. C. et al. (1997) FEBS Lett. 418, 139–143 31 Scheibel, T., Weikl, T. and Buchner, J. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 1495–1499 32 Tamura, Y. et al. (1997) Science 278, 117–120 33 Wearsch, P. A., Voglino, L. and Niccitta, C. V. (1998) Biochemistry 37, 5709–5719 34 Grenert, J. P. (1997) J. Biol. Chem. 272, 23843–23850 35 Smith, D. F. (1993) Mol. Endocrinol. 7, 1418–1429 36 Dittmar, K. D. et al. (1998) J. Biol. Chem. 273, 7358–7366 37 Kosano, H. et al. (1998) J. Biol. Chem. 273, 32973–32979 38 Cheetham, M. E. and Caplan, A. J. (1998) Cell Stress Chaperones 3, 28–36 39 Caplan, A. J. et al. (1995) J. Biol. Chem. 270, 5251–5257 40 Dey, B., Caplan, A. J. and Boschelli, F. (1996) Mol. Biol. Cell 7, 91–100 41 Kimura, Y., Yahara, I. and Lindquist, S. (1995) Science 268, Acknowledgements I am very grateful to Chris Nicchitta for the electron micrographs and to those colleagues who shared unpublished data and apologize to those whose work could not be cited owing to space constraints. Work in the author’s laboratory is supported by the NIH (DK49065) and the Irma T. Hirschl Trust. 267 reviews 42 43 44 45 46 47 48 49 50 51 52 53 54 1362–1365 Frydman, J. and Höhfeld, J. (1997) Trends Biochem. Sci. 22, 87–92 Johnson, B. D. et al. (1998) J. Biol. Chem. 273, 3679–3686 Prodromou, C. et al. (1999) EMBO J. 18, 754–762 Duina, A. A. et al. (1996) Science 274, 1713–1715 Dittmar, K. D. et al. (1997) J. Biol. Chem. 273, 7358–7366 Fang, Y. et al. (1998) Mol. Cell. Biol. 18, 3727–3734 Bohen, S. (1998) Mol. Cell. Biol. 18, 3330–3339 Knoblauch, R. and Garabedian, M. J. (1999) Mol. Cell. Biol. 19, 3748–3759 Fang, Y. et al. (1996) J. Biol. Chem. 271, 28697–28702 Young, J. C., Obermann, W. M. J. and Hartl, F. U. (1998) J. Biol. Chem. 273, 18007–18010 Carrello, A. et al. (1999) J. Biol. Chem. 274, 2682–2689 Owens-Grillo, J. K. et al. (1996) J. Biol. Chem. 271, 13468–13475 Dolinski, K., Cardenas, M. E. and Heitman, J. (1998) Mol. Cell. Biol. Occludin and claudins in tight-junction strands: leading or supporting players? Shoichiro Tsukita and Mikio Furuse Tight junctions have attracted much interest from cell biologists, especially electron microscopists, since on freeze–fracture electron microscopy they appear as a well-developed network of continuous, anastomosing intramembranous strands (tight-junction strands). These strands might be directly involved in the ‘barrier’ as well as ‘fence’ functions in epithelial and endothelial cell sheets, but until recently little was known of their constituents. This review discusses current understanding of the molecular architecture of tight-junction strands, focusing on the recent discovery of two distinct types of tightjunction-specific integral membrane proteins, occludin and claudins. Multicellular organisms contain various compositionally distinct fluid compartments, which are established by epithelial and endothelial cell sheets. For these cell sheets to function as barriers maintaining the distinct internal environment of 268 18, 7344–7352 55 Marsh, J. A., Kalton, H. M. and Gaber, R. F. (1998) Mol. Cell. Biol. 18, 7353–7359 56 Chang, H-C. J., Nathan, D. F. and Lindquist, S. (1997) Mol. Cell. Biol. 17, 318–325 57 Duina, A. A. et al. (1998) J. Biol. Chem. 273, 10819–10822 58 Silverstein, A. M. et al. (1998) J. Biol. Chem. 273, 20090–20095 59 Kimura, Y. et al. (1997) Genes Dev. 11, 1775–1785 60 Hunter, T. and Poon, Y. C. (1997) Trends Cell Biol. 7, 157–161 61 Fliss, A. E. et al. (1997) Mol. Biol. Cell 8, 2501–2509 62 Duina, A. A., Kalton, H. M. and Gaber, R. F. (1998) J. Biol. Chem. 273, 18974–18978 63 Rutherford, S. L. and Lindquist, S. (1998) Nature 396, 336–342 64 Garcia-Cardena, G. et al. (1998) Nature 392, 821–824 65 Holt, S. E. et al. (1999) Genes Dev. 13, 817–826 66 Höhfeld, J. (1998) Biol. Chem. 379, 269–274 each compartment, the paracellular pathway between adjacent cells in the sheet must be tightly sealed to prevent the diffusion of solutes, and the apical and basolateral membrane domains must be differentiated to allow active transport across the sheet. Tight junctions (TJs) are directly involved in paracellular sealing (barrier function) as well as in membrane domain differentiation (fence function) in epithelial and endothelial cell sheets1,2 (Fig. 1a). Tight junctions were first identified by ultrathinsection electron microscopy as a type of intercellular junction3. They appear as a series of discrete sites of apparent membrane fusion (so-called ‘kissing points’), involving the outer leaflet of the plasma membranes of adjacent cells. By freeze–fracture electron microscopy, TJs appear as a set of continuous, anastomosing strands in the P-face (the cytoplasmic leaflet of plasma membranes; see Box 1), with complementary grooves in the E-face (the extracytoplasmic leaflet)4 (Fig. 1b). These ultrastructural observations led to the structural model of TJs described in Fig. 1a. In this model, TJ strands associate laterally with those in apposing membranes, always forming paired strands, which eliminate the extracellular space between adjacent cells and function as a fence between apical and basolateral membrane domains. In the past few decades, a great deal of effort has been expended clarifying the molecular architecture of TJ strands as this is clearly a prerequisite for a better understanding of the physiological relevance of the barrier and fence functions of TJs2. An initial model was that some integral membrane proteins polymerize into strands within plasma membranes. A second model was that the TJ strand itself is predominantly lipid in nature – that is, formed of cylindrical inverted lipid micelles5–7. The detergent stability of TJ strands visualized by electron microscopy8 suggested that TJ strands are not solely lipid in composition, but to settle this debate we had to wait for the identification of occludin and claudins. 0962-8924/99/$ – see front matter © 1999 Elsevier Science. All rights reserved. PII: S0962-8924(99)01578-0 trends in CELL BIOLOGY (Vol. 9) July 1999