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