REVIEWS

Chaperone machines for protein

folding, unfolding and disaggregation
Helen Saibil
Abstract | Molecular chaperones are diverse families of multidomain proteins that have
evolved to assist nascent proteins to reach their native fold, protect subunits from heat shock
during the assembly of complexes, prevent protein aggregation or mediate targeted
unfolding and disassembly. Their increased expression in response to stress is a key factor in
the health of the cell and longevity of an organism. Unlike enzymes with their precise and
finely tuned active sites, chaperones are heavy-duty molecular machines that operate on a
wide range of substrates. The structural basis of their mechanism of action is being
unravelled (in particular for the heat shock proteins HSP60, HSP70, HSP90 and HSP100) and
typically involves massive displacements of 20–30 kDa domains over distances of 20–50 Å
and rotations of up to 100º.

Autophagy Protein quality control, also known as proteostasis, diverse proteins? Most of the main chaper­ones use cycles
A process in which intracellular constitutes the regulation of protein synthesis, folding, of ATP binding and hydrolysis to act on non-native poly-
material is enclosed in a unfolding and turnover. It is mediated by chaperone peptides, facilitating their folding or unfolding 6. Others
membrane compartment and and protease systems, together with cellular clearance simply have a ‘handover’ role, protecting nascent sub­
delivered to the lysosome
(vacuole in yeast) for
mechanisms such as autophagy and lysosomal degrada- units during the assembly process. Some ATP-dependent
degradation and recycling tion. These quality control systems have an essential chaperones, also known as protein remodelling factors,
of the macromolecular role in the life of cells, ensuring that proteins are cor- mediate targeted disassembly, unfolding or even rever-
constituents. rectly folded and functional at the right place and time1,2. sal of aggregation. Because of the disordered nature of
They are crucial for mitigating the deleterious effects of unfolded, partially folded or aggregated proteins, struc-
Unfolded protein response
(UPR). A signalling system that
protein misfolding and aggregation, which, by unclear tural details are lacking for the interactions between
regulates the balance between mechanisms, can cause cell death in neurodegeneration chaperones and their protein substrates.
folding capacity of the and other incurable protein misfolding diseases (BOX 1). An emerging functional feature of chaperones is their
endoplasmic reticulum (ER) A set of protein families termed molecular chaperones highly dynamic behaviour. Despite the great importance
and protein synthesis.
If misfolded proteins
assists various processes involving folding, unfold- and utility of X‑ray crystal structures, the resulting
accumulate, this pathway ing and homeostasis of cellular proteins. After protein atomic structures can give a misleading impression of
triggers apoptosis. denaturation caused by stress (for example, due to heat static, fixed conformations. It seems that the conforma-
or toxin exposure) or disease conditions, proteins can tions of these ATPases are only weakly coupled to their
Heat shock proteins
be unfolded, disaggregated and then refolded, or they nucleotide states (that is, whether they are bound to
(HSPs). The expression of these
proteins is greatly enhanced by
can be targeted for disposal by proteolytic systems. ATP, ADP or in the unbound state) and that they are in
increased temperature or other Found in all cellular compartments, chaperones act on a continua­l state of rapid fluctuation.
stress conditions. Most a broad range of non-native substrates. The endoplasmic This Review focusses on the roles and mechanisms
chaperones are HSPs. reticulum (ER), in particular, is a major site for protein of representatives of the major families of general, ATP-
production and quality control in membrane and secre- dependent chaperones, namely the heat shock proteins
Department of
tory systems. If it is overburdened by misfolded proteins, (HSPs; also known as stress proteins) HSP60, HSP70,
Crystallography, Institute for
Structural and Molecular the unfolded protein response (UPR) triggers cell death HSP90 and HSP100. We summarize our current under-
Biology, Birkbeck College by apoptosis3. standing of these allosteric machines and address the ways
London, UK. Chaperones are not typical macromolecular machines in which the energy of ATP binding and hydrolysis are
e-mail: [email protected]. with a well-defined substrate. The major molecular chap- used to unfold misfolded polypeptides for either refold-
bbk.ac.uk
doi:10.1038/nrm3658
erones (TABLE 1) have little specificity but provide essential ing or disaggregation. These considerations underlie a
Published online assistance to a complex and highly specific process, pro- key unanswered question: which protein conformations
12 September 2013 tein folding 4,5. How do they assist folding or unfolding of have chaperones evolved to prevent (under conditions

630 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

Box 1 | Protein misfolding diseases

HSP70 and HSP90 are highly interactive, functioning
with many partners and cofactors. Conversely, HSP60
Mutations that destabilize a protein can cause the loss of protein function. If the protein and HSP100 are ‘loners’. They have few interacting part-
is degraded and aggregation is prevented, serious pathological consequences may be ners and their active sites are not exposed on the outer
avoided. However, the aggregation of misfolded proteins creates toxicity (toxic gain of surface of the protein complex. Despite their very dif-
function). Simple loss-of-function mutations in CFTR (cystic fibrosis transmembrane
ferent modes of action, these general chaperones share
conductance regulator) destabilize the protein, leading to its misfolding in the
endoplasmic reticulum (ER) and subsequent degradation, but they do not cause cell
the common property of binding various non-native
death. Conversely, retinitis pigmentosa mutations in the highly abundant proteins to prevent their aggregation.
photoreceptor protein rhodopsin affects its folding and transport and eventually result
in photoreceptor cell death and blindness111,112. HSP70 — a tuneable chaperone system
Serious neurodegenerative conditions, including Alzheimer’s disease, Parkinson’s HSP70 is the most abundant chaperone and exists as
disease, Huntington’s disease and prion disease, result from the aggregation of a many orthologues in different cellular compartments.
diverse set of peptides and proteins associated with the conversion to amyloid-like In association with various cofactors it carries out diverse
fibrillar assemblies. Although neurodegenerative diseases present an obvious burden in functions, including protein folding, trans­location across
ageing societies, systemic conditions involving amyloids such as type II diabetes are organelle membranes and disaggregation of aggregates.
equally serious. The common structural feature of amyloid is its cross β-fold in which
HSP70 has two domains: an ATPase domain and a
the protein, whatever its native structure, is converted into a largely or wholly β-strand
form. Short strands stack into ribbons that wind into fibrils with the strands running
s­ubstrate-binding domain. Its activity depends on
perpendicular to the fibril axis113–115. dynamic interactions between these two domains and
Although the structural and mechanistic basis of cytotoxicity remain obscure, there is also on interactions between these domains and co-
evidence for membrane damage by oligomeric intermediates in amyloidogenesis, in chaperones such as the HSP40 proteins (also known
addition to overload of protein quality control systems. In healthy individuals, as J proteins, named after E. coli DnaJ) and nucleotide
chaperones prevent or rescue cells from pathological consequences by promoting exchange factors (NEFs, which stimulate ADP release
refolding, degradation or sequestration into non-toxic aggregates116–119. and nucleotide exchange after ATP hydrolysi­s)6–8.
The insulin-like signalling pathways that regulate lifespan provide a link between
ageing and loss of proteostasis capacity1,2. The role of chaperones in these processes Cellular functions of HSP70. Even transient binding of
has prompted efforts to chemically modulate these systems, with the goal of providing
an extended segment of a polypeptide chain to HSP70
global protection against protein misfolding120,121.
could prevent misfolding and aggregation and maintain
the substrate in an unfolded state for translocation to
of stress or in misfolding diseases), that is, what is the another cellular compartment. Indeed, the HSP70 sys-
nature of the cytotoxic species that result when protein tem is an important component of the organelle trans-
homeostasis fails? location system on both sides of the membrane. The
conformational cycle of HSP70 is used both for delivery
Chaperone families of the substrate protein to the translocase that transports
Members of the HSP60 (known as GroEL in Escherichia it across the organelle membrane and to capture or pull
coli), HSP70 (known as DnaK in E. coli), HSP90 (known on the translocated polypeptide (reviewed in REF. 9).
as HptG in E. coli) and HSP100 (known as ClpA and Regarding folding from the unfolded state, it seems
ClpB in E. coli) families (the number indicates the molec- likely that polypeptides can collapse into their native
ular mass of each HSP subunit) interact either with aggre- fold in free solution upon release from HSP70. Failure to
gation-prone, non-native polypeptides or with proteins reach the correctly folded state would lead to re-binding.
tagged for degradation. Thus, the role of HSP70 in folding seems to be stabilizing
HSP70 coordinates cellular functions by directing the unfolded state or unfolding proteins until they can
substrates for unfolding, disaggregation, refolding or deg- spontaneously fold upon reaching their correct cellular
radation. HSP90 integrates signalling functions, acting at destination10.
a late stage of folding of substrates that are important in In addition to its role in folding, HSP70 has other
cellular signalling and development and targeting sub- specific cellular functions. For example, together with
Allosteric machines
Macromolecular complexes in strates for proteolysis. By contrast, HSP60 acts at early auxilin (which is also a J protein co‑chaperone), it dis-
which the activity is indirectly stages of folding and provides an outstanding example of assembles the clathrin coat on membrane vesicles after
modulated by binding of an a highly coordinated and symmetric allosteric machine completion of clathrin-mediated endocytosis11. It also
effector at a site remote from for protein folding. HSP100 is a sequential ‘threading’ cooperates with HSP100 ATPases in disaggregating
the active site. This induces
shifts in the domain or subunit
machine for unfolding that cooperates with either a pro- large aggregates (see below). The corresponding part-
structure that influence the tease ring for degradation or HSP70 for disaggregation, ner of HSP70 for disaggregation and/or detoxification
conformation of the active site. thus avoiding the toxic effects of aggregation. of aggregates in the cytosol of higher eukaryotes has
The mechanisms of action and allostery of the HSP60 recently been identified as the NEF HSP110, which also
Amyloid
and HSP70 families are understood in some detail. HSP60 has chaperone activity 12–15.
Protein species that form
deposits consisting of fibrillar forms symmetrical, self-contained complexes in which
protein aggregates rich in the substrate- and nucleotide-binding sites are located Structural basis of HSP70 function. The atomic struc-
β-sheet structure. They inside cavities, and they act in a concerted and global way tures of the ATPase domain and the substrate-binding
assemble from proteins that on the substrate. By contrast, HSP70 exposes regulatory domain of HSP70 were determined separately in the
have unfolded or misfolded.
About 20 distinct protein
surfaces and cooperates with various binding proteins that 1990s. Unexpectedly, the ATPase domain was found to
species are associated with can redirect its activity. It acts locally on unfolded regions have the same fold as actin and hexokinase, with two
particular amyloid diseases. of the substrate polypeptide. flexible domains surrounding a deep, nucleotide-binding

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 631

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

Table 1 | ATP-dependent chaperones, examples of their cofactors and functions

Chaperones* Cofactors Functions
Chaperonins
HSP60 (also known as CPN60), HSP10 (also known as CPN10), Protein folding, prevention of aggregation
GroEL (Escherichia coli), GroES, prefoldin
CCT (mammals),
thermosome (archaea)
HSP70 system
DnaK (E. coli), HSP40, DnaJ, Sis1, Hdj1, NEFs, Unfolding, disaggregation, stabilization of extended chains,
Ssa, Ssb (Saccharomyces cerevisiae), GrpE, HSP110 translocation across organelle membranes, folding, regulation of the
BiP (also known as GRP78) (mammals; ER) heat-shock response, targeting substrates for degradation
HSP90 system
HptG (E. coli), HOP, p50, AHA1, p23, FKPB52, Binding, stabilization and maturation of steroid receptors and protein
GRP94 (ER) UNC45 kinases, delivery to proteases, buffer for genetic variation, regulation of
substrate selection and fate, myosin assembly
HSP100
ClpA, ClpB, ClpX, HslU (bacteria; HSP70 system, ClpP, ClpS Unfolding, proteolysis, thermotolerance, resolubilization of aggregates,
mitochondria and chloroplasts), remodelling
p97, RPT1–RPT6 (eukaryotic)
AHA1, activator of HSP90 ATPase 1; BiP, binding immunoglobulin protein; CCT, chaperonin-containing TCP1 complex; ER, endoplasmic reticulum; FKBP52, 52 kDa
FK506‑binding protein; GRP78, 78 kDa glucose-regulated protein; HSP, heat shock protein; HOP, HSC70–HSP90‑organizing protein; NEFs, nucleotide exchange
factors.*The species and/or localization is specified in brackets.

cleft that closes around ATP16,17 (FIG. 1a). The substrate- the nucleotide-binding cleft creates a binding site on the
binding domain is thin and brick-shaped with a cleft ATPase domain for the interdomain linker, which then
capped by a mobile α‑helical lid. Both the lid and the recruits the substrate-binding domain. This domain
cleft open to allow substrate binding, which can then be docking distorts the substrate-binding cleft and opens
trapped by the closing lid18 (FIG. 1a). The nucleotide state the lid, which then binds to a different part of the
of the ATPase domain affects the opening (stimulated ATPase domain20–22.
by ATP binding) and shutting (after ATP hydrolysis) of
the substrate-binding site. However, the two domains, Regulation by co-chaperones. HSP70 acts together
which are connected by a flexible linker, are not seen with two co-chaperones in protein folding, namely an
together in most crystal structures. The flexible linker, HSP40 and a NEF. The HSP40 family is very diverse,
located at the base of the two domains remote from the with many specialized members targeting HSP70
cleft openin­g, is a key site in allosteric regulation. to specific sites or functions 7. HSP40 is thought to
A first view of the domain interaction came from act as the primary substrate recruiter forHSP70 and
the structure of yeast Sse1 (a homologue of mamma- stimulates the HSP70 ATPase. For pathways involving
lian HSP110)19. Although Sse1 and HSP110 are struc- nonspecific protein folding and refolding, the general
tural homologues of HSP70, they act as HSP70 NEFs. HSP40 is an elongated, V‑shaped dimer containing
More recently, the crystal structure of a d­isulphide- the charac­teristic, helical J domain that activates the
trapped form of ATP-bound DnaK (which is the HSP70 ATPase by binding at or near the interdomain
E. coli homologue of mammalian HSP70), together linker 23,24. The J domain is followed by a disordered,
with methyl transvers­e relaxation optimized spectroscop­y Gly-Phe-rich region, two tandem β-subdomains and
(methyl TROSY) nuclear magnetic resonance (NMR) a dimerization domain25. One of the β-subdomains
and mutational probing of the domain association contains a surface-exposed substrate-binding site.
in a mutant deficient in ATP hydrolysis revealed It seems likely that hydrophobic segments of a sub-
how the HSP70 domains interact20–22. Unlike Sse1 or strate polypeptide, following their initial recruitment
HSP110, DnaK is highly dynamic, with rapid fluctua- to the shallow, accessible binding sites of HSP40, are
Methyl transverse
relaxation optimized
tions between the docked and free conformations in all delivered to the deeper channel of HSP70 for binding
spectroscopy nucleotide-bound states. via the polypeptide backbone, with the J domain stimu-
(methyl TROSY). A method that A remarkable feature of the domain-docked comple­x lating the ATPase26,27,8. Thus, J proteins interact with
uses selective isotope labelling is the intimate association of the substrate-binding both the nucleotide- and s­ubstrate-binding domains
of methyl groups on protein
domain with the ATPase domain. The substrate- of HSP70, with flexibly linked sites stimulating the
side chains with a transverse
relaxation scheme optimized bindin­g domain is almost turned inside-out to wrap ATPase and delivering the bound polypeptide. NEFs
for methyl groups to obtain around the ATPase domain, with an extremely open such as E. coli GrpE or eukary­otic HSP110 interact near
well-resolved nuclear magnetic orientation of its helical lid and a scissor-like motion the entrance to the nucleotide cleft, moving the HSP70
resonance (NMR) spectra from of its β-subdomain that opens up the peptide-binding subdomain IIb (FIG. 1a) and opening the cleft for nucleo­
large protein structures far
beyond the normal range
cleft (FIG.  1b). It has been proposed that the HSP70 tide exchange28–30. Although these inter­actions have
obtained in NMR structure mechanism of action involves several key steps. First, been observed separately, how the dynamic complex
determination. allosteric signalling from ATP binding and closure of function­s as a whole has not yet been shown.

632 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

a ADP-bound
Side view
IIb
b ATP-bound
Lid

ADP Substrate

ATP
Linker

Substrate-binding
domain

Nucleotide-binding domain

Substrate

ADP Lid ATP

Nucleotide-binding Substrate-binding
domain domain
Figure 1 | HSP70 assemblies.  a | In the ADP-bound or nucleotide-free state, the nucleotide-binding domain (green;
Protein Data Bank (PDB) code: 3HSC)16 of heat shock protein 70 (HSP70) is connectedNature
by Reviews Molecular
a flexible| linker to theCell Biology
substrate-binding domain (blue; PDB code: 1DKZ), with the lid domain (red) locking a peptide substrate (yellow) into the
binding pocket18. A side view of the substrate domain is shown on the right. A cartoon depicting the two-domain complex
is shown below. The bound nucleotide is shown in space filling format. b | In the ATP-bound state, the lid opens, and both
the lid and the substrate-binding domain dock to the nucleotide-binding domain (PDB code: 4B9Q)20. The corresponding
cartoon of this conformation is shown below. When ATP binds, the cleft closes, triggering a change on the outside of the
nucleotide-binding domain that creates a binding site for the linker region. Linker binding causes the substrate-binding
domain and the lid domain to bind different sites on the nucleotide-binding domain, resulting in a widely opened
substrate-binding site that enables rapid exchange of polypeptide substrates. After hydrolysis, the domains separate and
the lid closes over the bound substrate. Such binding and release of extended regions of polypeptide chain are thought
to unfold and stabilize non-native proteins either for correct folding or degradation.

HSP90 — a cellular signalling hub of such proteins could serve to improve fitness during
HSP90, another highly abundant and ubiquitous chape­ evolutionary change. Some experimenta­l support for
rone, has diverse biological roles, but its mechanism of this idea comes from studies investigating the devel-
action is less well-understood than that of the major opmental effects of HSP90 inhibitors on Drosophila
chaperones. It is a highly flexible, dynamic protein and, melano­g aster and Arabidopsis thaliana, and from
in eukaryotes, has a multitude of interactors that regu- studies examining the effects of environmental stress
late its activities, making it a hub for many pathways31–34. in yeast 36.
Like other stress proteins, HSP90 is capable of binding
non-native polypeptides and preventing their aggrega- HSP90 in complex with nucleotides and substrates.
tion. It seems to act mainly at the late stages of sub- HSP90 forms a dimer of elongated subunits, with
strate folding. For example, steroid hormone receptors each subunit comprising three domains that are
must bind HSP90 for efficient loading of their steroid linked by flexible regions. It stably dimerizes through
ligand. The bacterial form seems to act alone and is not its c­arboxy‑terminal domains and also transiently
crucial for viability, but the eukaryotic forms and their through its amino‑terminal ATPase domain when ATP
many co-chaperones are essential. HSP90 is function- is bound37 (FIG. 2). HSP90 is extremely dynamic, as it
ally more specialized than the other general chaperones. fluctuates rapidly between conformations ranging from
GHKL
An ATP-binding superfamily It is important for maturation of signalling proteins in an open V-shape to a closed form resembling a pair of
that includes DNA gyrase, the development and cell division, and its substrates include cupped hands38. The nucleotide-binding site accom-
molecular chaperone heat steroid hormone receptors, kinases and key oncogenic modates a bent conformation of ATP, the binding of
shock protein 90, the DNA- proteins such as the tumour suppressor p53. which causes transient dimerization of the N‑terminal
mismatch-repair enzyme MutL
and His kinase, which bind ATP
An intriguing evolutionary hypothesis proposes that domains, characteristics of the GHKL (gyrase, HSP90,
in a characteristic bent HSP90 acts as a buffer for genetic variation by rescuing His kinase and MutL) ATPase fold shared with the DNA-
conformation. mutated proteins with altered properties35. A reservoir unwinding enzyme DNA gyrase39,40. Specific inhibitors

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 633

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

a Open b ADP-bound

N-terminal

Middle

C-terminal

c ATP-bound d HSP90–p50–CDK4

Figure 2 | HSP90 conformations and substrate binding.  Crystal structures of heat shock protein 90 (HSP90) dimers in an
open, unliganded state (Protein Data Bank (PDB) code: 2IOQ)122 (part a), a partly closed, ADP-bound
Nature Reviews | state (PDB code:
Molecular Cell Biology
2O1V)123 (part b) and in a closed, ATP-bound state (PDB code: 2CG9)37 (part c), are shown, and the amino-terminal domain
(green), the middle domain (yellow) and the carboxy‑terminal domain (blue) are indcated. The open form shown is
Escherichia coli HptG, the partly closed ADP-bound form is the canine endoplasmic reticulum-associated HSP90
homologue GRP94 and the ATP-bound form (shown is the ATP analogue AMP-PNP) is yeast Hsc82 (heat shock cognate 82).
Nucleotides are shown in space filling format. ATP favours binding to the closed form (part c), whereas hydrolysis
or nucleotide release is favoured by a range of more open states (parts a,b). Opening and closing of the cleft are thought to
mediate the action of HSP90 on its substrates, although the mechanisms underlying HSP90 action remain largely unclear.
The electron microscopy map of HSP90 in complex with the cofactor p50 and its substrate cyclin-dependent kinase 4
(CDK4) is shown48 (part d). Extra density of the side of this asymmetric complex is attributed to the cofactor and substrate.

of the HSP90 ATPase have marked effects in develop- FK506‑binding protein (FKBP52), are involved in
ment and cance­r. Although the HSP90 nucleotide state complexes with steroid receptors. These co-chaperones
is only weakly coupled to its conformational change41, interact through their tetratricopeptide repeat (TPR)
the many binding partners of HSP90 influence dif- domains with a conserved C‑terminal motif found on
ferent steps in the functional cycle. HSP90 action is HSP90 and also on HSP70 (REF. 43). Numerous other
modulated by co-chaperone­s and client proteins (the co-chaperone complexes assemble on HSP90 via TPR
term used for ‘substrates’ in the HSP90 system). In domains. For example, HSC70–HSP90‑organizing
addition, phosphoryl­ation, acetylation and other post- protein (HOP; also known as STI1) recruits HSP70 to
translationa­l modifications affect its functional state32. HSP90, creating a complex for substrate handover44.
Co-chaperones that target HSP90 to specific types Important non-TPR containing co-chaperones include
of client protein include p50 (also known as CDC37), activator of HSP90 ATPase 1 (AHA1) and p23, which is
which recruits kinases and inhibits the ATPase activity involved in client protein maturation45,37. Together with
of HSP90 (REF. 42). A set of co-chaperones with prolyl HSP70, HSP90 also has an important role in targeting
isomerase activity, such as the immunophilin 52 kDa substrates for degradation46.

634 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

Details of substrate binding to HSP90 are poorly This culminates in a power stroke that ejects the sub-
understood. Evidence for substrate-binding sites on all strate from the hydrophobic sites and simultaneously
three domains of HSP90 came from low-resolution elec- traps it inside the GroES-capped hydrophilic chamber
tron microscopy and mutational studies, which led to a for folding 62. Once encapsulated, the lack of exposed
model of hydrophobic surfaces lining the cavity of an hydrophobic sites or other partners for aggregation,
open dimer 47–50. Despite the lack of a mechanistic under- together with the limited enclosure (~7 nm maximum
standing of the action of HSP90, specific inhibitors of its dimension), blocks further misfolding or aggregation
ATPase activity, such as geldanamycin, were shown to pathways, so that the substrate can either follow a fold-
have important biological effects and form the basis for ing pathway determined by its amino acid sequence or
successful anticancer drugs51. remain unfolded. After a slow ATP hydrolysis step, the
chamber is re-opened, releasing the protein either com-
HSP60 — a protein folding container mitted to final folding and assembly or releasing it in a
Chaperonins (a term specific to this chaperone family) non-native state that will be recaptured by a chaperonin
can be divided into two subfamilies: group I is composed ring. For substrates that are too large to be encapsulated,
of the bacterial chaperonin GroEL and its co‑chaperonin GroES may still act allosterically to effect productive
GroES, as well as the mitochondrion- and chloroplast- release of the substrate from the remote open ring 63.
specific HSP60 proteins together with their HSP10 co‑ Key to understanding this action is to determine the
chaperonins; and group II chaperonins, which are found structures of the intermediate complexes when substrate
in archaea and the eukaryotic cytosol and comprise the and ATP have bound and GroES is being recruited. At
archaeal thermosome and eukaryotic CCT (chaperonin- low to intermediate resolution, substrate binding and
containing TCP1; also known as TriC). In group II, an GroEL domain movements have been characterized by
extra protein domain replaces the group I co‑chaperoni­n. single particle cryo-electron microscopy (cryo-EM) of
The bacterial GroEL–GroES chaperonin syste­m is by far various intermediate complexes, using statistical analy-
the best understood general chaperone. sis to discriminate multiple three-dimensional structures
Chaperonins are self-contained machines that leave from images of heterogeneous and dynamic complexes.
little to chance; they provide a complete isolation cham- This approach has yielded structural descriptions of
ber for protein folding. Early work on bacteriophage chaperonin complexes at different stages of substrate
assembly, mitochondrial and chloroplast biogenesis binding and folding and has enabled analysis of the
led to the realization that related proteins in bacteria, allosteric machinery 60,64,65.
chloroplasts and mitochondria have an essential role in Crystal structures as well as kinetic and mutational
de novo protein folding and assembly as well as refolding studies have revealed key allosteric sites in chaperonins.
stress-denatured proteins52–54. Each subunit contains three domains connected by flex-
ible hinge points (FIG. 3b). The nucleotide-binding pocket
Chaperonin structures and action. Biochemical, bio- is in the equatorial domain, and helix D runs from an
physical and structural analyses, particularly of E. coli Asp residue coordinating the γ‑phosphate site to one of
GroEL–GroES, have revealed many important parts of the two inter-ring contacts. In the GroES-bound state, the
the mechanism of action55,56. GroEL crystal structures intermediate domain closes over the ATP pocket, bringing
reveal details of the start and end states of extensive a catalytic Asp residue close to the nucleotide. Within each
movements of this chaperonin through concerted rigid- ring, the subunits are interlinked by salt bridges and act
body rotations of the subunit domains. Unliganded (apo) in concert, exhibiting positive cooperativity for ATP bind-
GroEL forms a 15 nm long cylindrical structure com- ing66. Conversely, the two rings act sequentially, exhibiting
posed of back‑to‑back rings of seven 60 kDa subunits57 negative cooperativity, which is transmitted through the
(FIG. 3a). These rings surround open cavities of ~5 nm two inter-ring contacts. Hydrophobic sites on the apica­l
diameter, the walls of which are lined by a band of contin- domain form the GroES- and substrate-binding sites
uous hydrophobic surfaces. The two rings alternately go (FIG. 3c). A mobile loop of GroES binds to the distal part
through cycles of ATP binding and hydrolysis. Upon ATP of this site, a region also implicated in substrate binding,
binding, a GroEL ring rapidly recruits the co‑chaperoni­n leading to the notion that GroES and substrate binding
GroES, a ring of seven 10 kDa subunits, which caps the are mutually exclusive. However, there is biochemical
cavity, entailing a dramatic structural reorganization to evidence, although no direct structural information, for
convert the open ring into an enclosed chamber with an intermediate state in which GroES and substrate are
a hydrophilic lining 58. In the GroES-bound ring, the simultaneously bound to GroEL67,68.
substrat­e-binding apical domains are elevated by 60º and
twisted by 90º relative to the unliganded ring. Substrate complexes. Crystal structures of extended
The open, hydrophobic lined ring is the acceptor peptides bound to the GroEL apical domain occupy
state that captures non-native polypeptides with exposed the same site as the GroES mobile loop58,69,70. This pro-
hydrophobic surfaces, accounting for the lack of bind- vides a partial view of how substrates might bind, but
ing specificity of group I chaperonins. The interaction electron microscopy studies of GroEL with captured
with the substrate can extend over 3–4 adjacent GroEL non-native proteins show a preference for binding
subunits59,60. The actions of GroEL, ATP and GroES deeper inside the cavity in the more proximal part of
exert mechanical forces on the substrate that potentially the hydrophobic site60,64. Moreover, electron microscopy
result in unfolding of trapped, misfolded proteins 61. structures show how substrates bind to the open ring

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 635

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

a GroEL–GroES complex b

Apical
GroEL
Hinge 2

Intermediate

Hydrophobic
binding sites
Hinge 1
Equatorial

Contact 2
Contact 1 Helix D

c GroES d
Side view
GroEL

Newly folded
gp23

Top view

Non-native
gp23
Polar residue Non-polar residue

Figure 3 | GroEL conformations and substrate complexes.  a | Overview of unliganded (apo) GroEL (Protein Data Bank
(PDB) code:1OEL)57 (left) and the GroEL–GroES complex (PDB code: 1SVT)58 (right). Nature Reviews
The overall | Molecular
shapes Cell
are shown asBiology
blue
surfaces, with three subunits coloured by domain in red, green and yellow in apo GroEL. One subunit of GroEL and one of
GroES (cyan) are highlighted in the GroEL–GroES complex. b | Conformation of a GroEL subunit in the apo form (left) and
the GroES-bound form (right), with GroEL key sites indicated (GroES is not shown). c | Cartoons of complexes with folding
proteins. Hydrophobic surfaces and residues are shown in yellow and polar residues in green. d | Cut open view of the
cryo-electron microscopy structure (Electron Microscopy Data Bank code: EMD-1548) of GroEL (PDB code: 1AON) in
complex with bacteriophage 56 kDa capsid protein (gp31) (PDB code: 1G31), with a non-native gp23 (PDB code: 1YUE)
bound to both rings64. The pink density in the folding chamber corresponds to newly folded gp23, and the yellow density
in the open ring is part of a non-native gp23 subunit. The corresponding atomic structures are shown embedded in the
electron microscopy density map, except for the non-native substrate, which is unknown and only partially visualized
owing to disorder. The open ring with its hydrophobic lining is the acceptor state for non-native polypeptides, and binding
to multiple sites may facilitate unfolding. ATP and GroES binding to the chaperonin create a protected chamber with a
hydrophilic lining that allows the encapsulated protein to fold.

and how they appear in the folding chamber (FIG. 3d). ATP complexes and domain movements. How does
This enclosure imposes an upper limit of under 60 kDa the binding of ATP detach a non-native protein mul-
for protein subunits that can be encapsulated. To accom- tivalently bound on the hydrophobic surface, resulting
modate its 56 kDa capsid protein, gp23, bacteriophage in a free subunit isolated in the folding chamber? The
T4 encodes its own GroES homologue, gp31, to make conformation of open GroEL rings in the presence of
the cage slightly taller 71. The newly folded large domain ATP is extremely dynamic. Sorting of hetero­geneous
of gp23, encapsulated and trapped by using a non- complexes by single particle electron microscopy has
hydrolysabl­e ATP analogue, fills the chamber and dis- resolved a set of intermediate states that seem to be in
torts it 64. A trapped, non-native state of another large equilibrium until a ring is captured by GroES65. ATP
substrate protein, RuBisCo (r­ibulose‑1,5‑bisphosphat­e binding causes small movements of the equatorial
carboxylase oxygenase), has been visualized by cryo-EM, domains that are relayed both within and between rings.
revealin­g contacts to apical and equatorial domains72. In the ATP-bound ring, the movements are amplified

636 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

into large rotations of the apical domains that culmi- a substitute for GroES in capping the ring 75. The various
nate in dramatic reorganization of the substrate-binding archaeal forms usually have eightfold or ninefold sym-
surface. The movements involve a rotation about the metry, and eukaryotic CCT has eight related but distinct
equatorial–intermediate hinge, bringing the catalytic gene products forming the eight subunits of each ring.
Asp residue near the nucleotide-binding pocket. This CCT in particular has been very difficult to study, and
rotation leads to the breakage of two intersubunit salt even the order of subunits in a ring is controversial76,77.
bridges and transient generation of two new ones. In Unlike the archaeal forms and most other chaperones,
addition, the ATP-triggered domain movements are CCT does not seem to be a HSP. CCT has specialized
relayed through helix D (FIG. 3b) to the opposite ring subunits, with some binding known substrates such
via distortion of the inter-ring interface, thus mediating as actin and tubulin. Various open, intermediate and
negative cooperativity 73. closed conformations of intact group II complexes have
The recently solved crystal structure of a GroEL been described by X‑ray crystallography and cryo-EM
double mutant lacking two key salt bridges reveals a (for example, REFS 78–81). An interesting difference in
remarkable, asymmetric ring with ADP bound to every how the allosteric machinery operates is that the inter-
subunit 74. The seven different subunit conformations ring interface is formed of 1:1 instead of 1:2 subunit con-
correspond to those seen in the individual cryo-EM tacts, leading to altered allosteric interactions82,83. The
reconstructions65. In the cryo-EM structures, the rings unliganded form is open and dynamic, equivalent to
were observed to maintain sevenfold symmetry, except the open state of GroEL. ATP analogue binding seems
for the apical domains in the more open states. to gradually close the cage. Remarkably, although the
The observed arrangements of the substrate-binding apical domains undergo similar elevations and twists
surface fall into four categories (Supplementary infor- in group I and II chaperonins, these motions seem to
mation S1 (figure)). The distal part of the hydrophobic occur in reversed sequence (Supplementary informa-
site is delineated by helix H and helix I. The collinear tion S3 (movie)). Overall, it seems likely that group II
tracks of both helices lining the apo GroEL ring are dis- chaperonins perform similar actions as members of the
torted into tilted tracks with the end of helix H joined group I family, but they exhibit different ATP-driven
to the next helix I in one category of GroEL–ATP states. allosteric movements.
In these structures, the hydrophobic sites form a con-
tinuous band. In the more open GroEL–ATP states, HSP100 disassembly machines
the contacts between adjacent apical domains are com- The HSP100 proteins are unfoldases and disaggregases,
pletely lost, and the hydrophobic band becomes discon- forceful unfolding motors that deliver substrates to
tinuous. The free apical domains are not constrained to compartmentalized proteases or disassemble aggregates
remain in symmetric positions. The open state has two containing misfolded proteins.
important properties. First, radial expansion provides
a plausible mechanism for forced unfolding of multi- The AAA+ chaperones. HSP100 proteins are members
valently bound substrate. Second, combined with ring of the AAA+ superfamily, which typically form oligo-
expansion, the elevation of the helix H–helix I groove meric ring structures and have mechanical actions such
creates a suitable docking site for the GroES mobile as threading polypeptides or polynucleotides through a
loops. In order to reach the folding-active, GroES- central channel in order to unfold or unwind them84,85.
bound conformation, the GroES binding sites must each AAA+ proteins function in various cellular processes,
twist by 100º (Supplementary information S2 (movie)). including the disassembly of complexes, for example
Thus, the open state is a good candidate for the ini- the SNARE complexes that bring membranes together
tial GroES-docked intermediate: the mobile loops are for vesicle fusion. The role of chaperone members of
highly flexible and can easily be modelled without the this family is best characterized in regulated proteolysi­s.
twist they adopt in the GroEL–GroES crystal structures. At the core of these compartmentalized proteases is a
Moreover, the key parts of the substrate-binding site, stack of co-axial ATPase and protease rings, formed
helix I and the more proximal, underlying segment, are either by separate functional domains of a single sub­
still exposed to the cavity. It has been proposed that a unit type (as in the bacterial Lon protease) or in separate
ternary complex between the open state GroEL–ATP, ATPase and protease subunit rings (as in the HslUV (also
substrate and GroES represents the elusive intermediate, known as ClpYQ) complex)86 (FIG. 4a,b). In HslUV, both
and that the 100º twist, triggered by binding of the rings are hexameric, whereas others such as ClpAP have
GroES loops to produce the final bullet complex, would a symmetry mismatch with hexameric ClpA ATPase
provide the power stroke that removes the hydrophobic and heptameric ClpP protease rings87. Although the
binding site from the cavity and forcefully ejects the eukaryotic proteasome is much more complex, it has
bound substrate into the chamber for folding 65. the same core architecture, and its regulatory cap con-
tains a heterohexamer of ATPase subunits (RPT1–
Group II chaperonins. Group II chaperonins perform RPT6) that perform­s the same unfoldin­g and threading
similar functions to group I chaperonins, and the under- functions88,89.
lying machinery is closely related. The most obvious The defining feature of the superfamily is the AAA+
structural difference between group I and group II domain, which consists of an α–β subdomain and a
chaperonins is the presence of a prominent insertion in smaller, helical subdomain85 (FIG. 4c,d). The nucleotide-
the apical domain in group I chaperonins, which acts as binding site is located at the subdomain interface.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 637

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

a b

ClpA, ClpB Substrate

AAA+ domain
ClpX, HslU
ATPase
c
Protease

ATPase

Figure 4 | HSP100 unfoldase.  a | The two types of heat shock protein 100 (HSP100) sequences are shown schematically,
with either a single or two tandem AAA+ domains. The characteristic Walker A andNature
B sitesReviews | Molecular
are shown in red. b | Cell Biology
The HslUV
ATPase–protease complex is shown as a cartoon on the left, and the atomic structure is shown on the right (Protein Data
Bank (PDB) code: 1G3I)124. c | Top view of the asymmetric ClpX crystal structure (PDB code: 3HWS)95. The four bound ADP
molecules are shown in space-filling format and Tyr side chains on the pore loops are shown as magenta sticks. d | Side view
section of ClpX showing the pore with three of the Tyr sites at different heights.

Conserved regions important in nucleotide binding to spool the unfolding chain through the channel.
and hydrolysis are the Walker A and Walker B motifs, The structure of an asymmetric ClpX ring shows a
sensor 1 and sensor 2 as well as the Arg finger involved sequence of pore loops at different heights in the channel
in catalysis of the ATPase at the interface between sub­ and suggests a sequential or random action of the
units. AAA+ chaperones typically form hexameric rings sub­units around the ring 95 (FIG. 4d). Their axial sepa-
that surround a narrow central pore lined with loops ration of ~1 nm fits well with results obtained from
containing a substrate interaction site with aromatic single-molecul­e optical tweezer experiments showing
and hydrophobic side chains. They exist as both single translocation steps in multiples of 1 nm96. Force and
AAA+ rings (such as in HslU and ClpX) and stacked extension measurements support the action of a power
rings of tandem AAA+ domains (such as in ClpA, stroke rather than a ratchet mechanism capturing ran-
ClpB and ClpC). The HSP100 chaperones also have dom Brownian motions. The single-molecule approach
very mobile N‑terminal domains that can play a part shows that a C‑terminal subdomain of GFP is extracted
in substrate delivery to the central channel or interact first, and this destabilizes the rest of the β-barrel, which
with cofactors90,91. unfolds before it is delivered to the surface of ClpX.
Thus, for GFP, only the first unfolding step requires
Unfolding during ATP-dependent proteolysis. How does forceful pulling.
unfolding work? First, the substrate is targeted to the The AAA+ protein p97 (also known as CDC48 or
entrance of the HSP100 channel. In bacteria, ribosome VCP) functions in the transport of substrates to the
stalling causes expressed polypeptides to be marked for proteasome, in particular of proteins that are misfolded
degradation by addition of an 11‑residue peptide, the in the ER and are retrotranslocated to the cytosol for
small, stable 10S RNA ssrA tag, which targets them to degradation97,98. p97 is a highly conserved protein with
ClpXP or ClpAP92. Both ClpX and ClpA are powerful tandem AAA+ domains and a mobile N-terminal
unfoldases that can even rapidly unfold a stable protein domain and has recently been suggested to represent
like GFP, if it is suitably tagged93. The central channel is the ancestral proteasome unfoldase ring 99. p97, together
lined with Tyr residues on mobile pore loops that provide with cofactors, has various other roles when it is in
the binding sites for translocating chains, without speci- close proximity to membranes. These functions relate
ficity for sequence or chain polarity 94 (FIG. 4c,d). Once a more to the actions of family members such as NSF
polypeptide terminus or loop is engaged in the chan- (N-ethylmaleimide-sensitive factor), which disassem-
nel, rotations of the AAA+ subdomains, fuelled by the ble SNARE complexes at membrane surfaces after they
ATPase cycle, are thought to produce a rowing motion have mediated vesicle fusion.

638 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

Substrate

a b

N domain

Motif 2

AAA+ 1
Coiled
coil
HSP70

AAA+ 2

ClpB subunit

Figure 5 | HSP100–HSP70 disaggregase.  The crystal structure of a ClpB subunit (Protein Data Bank (PDB) code: 1QVR)103
(part a) and a schematic representation of the three-tiered hexamer are shown, with one ClpB coiled-coil domain
Natureshown
(dark blue) bound to heat shock protein 70 (HSP70; with the nucleotide-binding domain Reviews | Molecular
in green Cell Biology
and the
substrate-binding domain in blue (PDB code: 4B9Q)) (part b). The ClpB–Hsp70 complex is derived from the model in
REF. 106 combined with the structure of domain-docked HSP70 from REF. 20. The motif 2 sequence in the coiled-coil
domain is highlighted in pink. A substrate polypeptide (yellow) is being extracted from an aggregate and threaded
through the ClpB channel.

Protein disaggregation. A subset of the HSP100 chaper- tangentially on the surface109. However, ClpC lacks the
ones found in bacteria, plants and fungi have the unique HSP70‑binding arm of the coil and requires the co­factor
ability to reverse protein aggregation, in cooperation MecA for hexamer assembly. Recent work probing
with their cognate HSP70 system100–102. This subfamily accessibility and hydrogen–deuteriu­m exchange on the
includes E. coli ClpB and yeast Hsp104, which have tan- coiled-coil domain of E. coli ClpB does not support either
dem ATPase domains. A 90 Å long coiled-coil pro­peller, model. Rather, it suggests that the coil lies on the surface
inserted near the end of the first ATPase domain103, of the hexamer, with motif 2 being protected when ClpB
couples their unfolding and translocation actions to activity is repressed and being accessible when ClpB is
HSP70 (REF. 104) (FIG. 5a). Binding of the HSP70 ATPase active110 (FIG. 5b).
domain to one end of the coiled-coil, a region highly Methyl TROSY NMR has recently been used to
sensitive to mutations and known as motif 2, is required model the local interactions between the ClpB coiled-
for disaggregation105,106. Docking to low-resolution and coil and the DnaK ATPase domain in its open, ADP-
symmetrized electron microscopy maps yielded contro- bound state106. Combining this model with the model of
versial results regarding the hexamer arrangement and domain-docked DnaK suggests how DnaK might deliver
the degree of expansion of the ring. A model based on a polypeptide segment to ClpB (FIG. 5b). This specula-
studies of Thermus thermophilus ClpB proposes that tive, combined model suggests that the ClpB coiled-coil
the subunits are tightly packed around a 15 Å channel adopts a more vertical orientation to bring the DnaK
and the coiled-coils protrude as radial spikes103. By con- substrate-binding domain to the vicinity of the pore
trast, electron microscopy maps of yeast Hsp104 sug- channel. The N‑terminal domains of ClpB might play a
gest a much more expanded ring with a wide channel part in delivering the substrate from DnaK to ClpB, after
and the coiled-coils intercalated between the subunits, DnaJ makes initial weak contact with the surface of the
partly buried and partly exposed on the surface, with the aggregate and hands over a segment of the polypeptide
HSP70-binding tip of the coil adjacent to the N-terminal for engagement with DnaK.
ring 107. More recent cryo-EM maps of HSP104 are inter-
preted as typical AAA+ rings with the coiled-coils on the Conclusions
outside, but no density is observed for the coiled-coils108. Chaperones are nanoscale molecular machines that rec-
A low-resolution crystal structure of the hexa­meric ognize incompletely or incorrectly folded proteins, arrest
assembly of ClpC, a protease-coupled HSP100 with or unfold them and then either release them for sponta-
tandem AAA+ domains and a partial coiled-coil struc- neous refolding or target them for degradation. With the
ture, shows an expanded ring and the coiled-coil lying help of many cofactors, the general purpose chaperone

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 639

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

HSP70, a two-domain monomer, carries out all these ATPase in HSP100. HSP70 shares its nucleotide-binding
actions. Another ‘sociable’ chaperone, HSP90, acts as a fold with actin and hexokinase, whereas HSP90 has a
flexible dimer, with even more partners to regulate its GHKL nucleotide-binding fold characteristic of DNA
activities. In a more solitary action, the HSP60 chaper- gyrase. The nucleotide binds in an extended conforma-
onins assist folding by creating an isolation chamber for tion to HSP60 and HSP70 but is bent when bound to
the substrate protein. Most forceful of all, the HSP100 HSP90 and HSP100, giving rise to different specificities
protein remodellers can rip apart even stably folded for nucleotide analogues (Supplementary information
proteins or disassemble large and otherwise irreversible S4 (figure)).
aggregates. Although the chaperone systems discussed here have
HSP70 and HSP90 have many surface exposed inter- a fairly broad range of substrates, many proteins have
action sites for cofactors, giving them a high degree of specific requirements for chaperones and co‑chaperones.
regulation and integration into other cellular pathways. For example, the substrates of group I and group II chap-
By contrast, the HSP60 and HSP100 families are largely eronins are quite distinct; specific HSP40 co‑chaperones
inward looking, and they enclose their active sites with are required together with HSP70 for the folding of many
few cofactors. Their activities are mainly regulated by important substrates. The mechanisms of this specific-
a stress-induced increase in their expression levels. ity are poorly understood. A major current question is
A striking feature of the ATPase cycles of these chaper- why the chaperone systems become less effective in age-
ones is their highly dynamic nature. Rather than simple ing organisms, leading to the eventual failure of protein
conformational switching, the massive domain move- quality control and the onset of misfolding diseases.
ments in chaperone action are only loosely coupled to Future progress in the field will require high-resolution
their nucleotide-bound state. Nevertheless, each of these structures of chaperone complexes acting on misfolded
chaperone families has a distinct mode of ATP binding, or unfolded proteins, the identification of specific causal
ranging from the unique chaperonin nucleotide site pathways in aggregate and amyloid toxicity, as well as a
to the very widespread Walker A and Walker B type better understanding of the regulation of proteostasis.

1. Gidalevitz, T., Prahlad, V. & Morimoto, R. I. The stress 14. Song, Y. et al. Molecular chaperone Hsp110 rescues a 24. Ahmad, A. et al. Heat shock protein 70 kDa
of protein misfolding: from single cells to multicellular vesicle transport defect produced by an ALS- chaperone/DnaJ cochaperone complex employs an
organisms. Cold Spring Harb. Perspect. Biol. 3, associated mutant SOD1 protein in squid axoplasm. unusual dynamic interface. Proc. Natl Acad. Sci. USA
a009704 (2011). Proc. Natl Acad. Sci. USA 110, 5428–5433 (2013). 108, 18966–18971 (2011).
2. Taylor, R. C. & Dillin, A. Aging as an event of Reveals that inhibition of axonal transport by 25. Li, J. Z., Qian, X. G. & Sha, B. The crystal structure of
proteostasis collapse. Cold Spring Harb. Perspect. misfolded, disease-associated superoxide the yeast Hsp40 Ydj1 complexed with its peptide
Biol. 3, a004440 (2011). dismutase is reversed by the HSP70–HSP110 substrate. Structure 11, 1475–1483 (2003).
3. Korennykh, A. & Walter, P. Structural basis of the disaggregation machinery. 26. Rodriguez, F. et al. Molecular basis for regulation of
unfolded protein response. Annu. Rev. Cell Dev. Biol. 15. Mattoo, R. U. H., Sharma, S. K., Priya, S., Finka, A. & the heat shock transcription factor σ32 by the DnaK
28, 251–277 (2012). Goloubinoff, P. Hsp110 is a bona fide chaperone using and DnaJ chaperones. Mol. Cell 32, 347–358
4. Hartl, F. U., Bracher, A. & Hayer-Hartl, M. Molecular ATP to unfold stable misfolded polypeptides and (2008).
chaperones in protein folding and proteostasis. reciprocally collaborate with Hsp70 to solubilize 27. Petrova, K., Oyadomari, S., Hendershot, L. M.
Nature 475, 324–332 (2011). protein aggregates. J. Biol. Chem. 288, & Ron, D. Regulated association of misfolded
5. Bukau, B., Weissman, J. & Horwich, A. Molecular 21399–21411 (2013). endoplasmic reticulum lumenal proteins with p58/
chaperones and protein quality control. Cell 125, 16. Flaherty, K. M., Deluca-Flaherty, C. & McKay, D. B. DNAJc3. EMBO J. 27, 2862–2872 (2008).
443–451 (2006). 3‑dimensional structure of the ATPase fragment of a 28. Harrison, C. J., Hayer-Hartl, M., Di Liberto, M.,
6. Mayer, M. P. Gymnastics of molecular chaperones. 70 k heat-shock cognate protein. Nature 346, Hartl, F. & Kuriyan, J. Crystal structure of the
Mol. Cell 39, 321–331 (2010). 623–628 (1990). nucleotide exchange factor GrpE bound to the ATPase
7. Kampinga, H. H. & Craig, E. A. The HSP70 chaperone 17. Bork, P., Sander, C. & Valencia, A. An ATPase domain domain of the molecular chaperone DnaK. Science
machinery: J proteins as drivers of functional common to prokaryotic cell-cycle proteins, sugar 276, 431–435 (1997).
specificity. Nature Rev. Mol Cell. Biol. 11, 579–592 kinases, actin, and Hsp70 heat-shock proteins. 29. Schuermann, J. P. et al. Structure of the
(2010). Proc. Natl Acad. Sci. USA 89, 7290–7294 (1992). Hsp110:Hsc70 nucleotide exchange machine.
8. Zuiderweg, E. R. et al. Allostery in the Hsp70 18. Zhu, X. T. et al. Structural analysis of substrate binding Mol. Cell 31, 232–243 (2008).
chaperone proteins. Top. Curr. Chem. 328, 99–153 by the molecular chaperone DnaK. Science 272, 30. Polier, S., Dragovic, Z., Hartl, F. U. & Bracher, A.
(2013). 1606–1614 (1996). structural basis for the cooperation of Hsp70 and
9. Young, J. C. Mechanisms of the Hsp70 chaperone 19. Liu, Q. & Hendrickson, W. A. Insights into Hsp70 Hsp110 chaperones in protein folding. Cell 133,
system. Biochem. Cell Biol. 88, 291–300 (2010). chaperone activity from a crystal structure of the yeast 1068–1079 (2008).
10. Sharma, S. K., De los Rios, P., Christen, P., Lustig, A. Hsp110 Sse1. Cell 131, 106–120 (2007). 31. Pratt, W. B. & Toft, D. O. Steroid receptor interactions
& Goloubinoff, P. The kinetic parameters and energy 20. Kityk, R., Kopp, J., Sinning, I. & Mayer, M. P. with heat shock protein and immunophilin
cost of the Hsp70 chaperone as a polypeptide Structure and dynamics of the ATP-bound open chaperones. Endrocrine Rev. 18, 306–360 (1997).
unfoldase. Nature Chem. Biol. 6, 914–920 conformation of Hsp70 chaperones. Mol. Cell. 48, 32. Li, J., Soroka, J. & Buchner, J. The Hsp90 chaperone
(2010). 863–874 (2012). machinery: conformational dynamics and regulation
11. Rothnie, A., Clarke, A. R., Kuzmic, P., Cameron, A. & Obtains the crystal structure of the allosterically by co‑chaperones. Biochim. Biophys. Acta 1823,
Smith, C. J. A sequential mechanism for clathrin cage active, domain-docked form of HSP70 by disulphide 624–635 (2012).
disassembly by 70‑kDa heat-shock cognate protein stabilization of its ATP-bound conformation. Reveals 33. Didenko, T., Duarte, A. M. S., Karagoz, G. E. &
(Hsc70) and auxilin. Proc. Natl Acad. Sci. USA 108, a wide opening of the substrate-binding site. Rudiger, S. G. D. Hsp90 structure and function studied
6927–6932 (2011). 21. Zhuravleva, A., Clerico, E. M. & Gierasch, L. M. by NMR spectroscopy. Biochim. Biophys. Acta 1823,
12. Shorter, J. The mammalian disaggregase machinery: An interdomain energetic tug‑of‑war creates the 636–647 (2012).
Hsp110 synergizes with Hsp70 and Hsp40 to catalyze allosterically active state in Hsp70 molecular 34. Johnson, J. L. Evolution and function of diverse Hsp90
protein disaggregation and reactivation in a cell-free chaperones. Cell 151, 1296–1307 (2012). homologs and cochaperone proteins. Biochim.
system. PLoS ONE 6, e26319 (2011). Shows, using methyl TROSY NMR of an HSP70 Biophys. Acta 1823, 607–613 (2012).
Shows, together with reference 13 and 14, that mutant defective in ATP hydrolysis, the dynamics of 35. Taipale, M., Jarosz, D. F. & Lindquist, S. HSP90 at the
HSP110, in cooperation with the HSP70 system, the domain interactions in the formation of hub of protein homeostasis: emerging mechanistic
exerts the disaggregase activity in metazoan cells. alternative interfaces that are induced by the insights. Nature Rev. Mol. Cell Biol. 11, 515–528
13. Rampelt, H. et al. Metazoan Hsp70 machines interdomain linker region. (2010).
use Hsp110 to power protein disaggregation. 22. Qi, R. et al. Allosteric opening of the polypeptide- 36. Jarosz, D. F. & Lindquist, S. Hsp90 and environmental
EMBO J. 31, 4221–4235 (2012). binding site when an Hsp70 binds ATP. Nature Struct. stress transform the adaptive value of natural genetic
Shows that the HSP70 nucleotide exchange activity Mol. Biol. 20, 900–907 (2013). variation, Science 330, 1820–1824 (2010).
of HSP110 is essential for its role in 23. Jiang, J. et al. Structural basis of J cochaperone 37. Ali, M. M. et al. Crystal structure of an Hsp90–
disaggregation, whereas HSP110 ATPase and binding and regulation of Hsp70. Mol. Cell 28, nucleotide–p23/Sba1 closed chaperone complex.
intrinsic chaperone activity are not required. 422–433 (2007). Nature 440, 1013–1017 (2006).

640 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved

 F O C U S O N q ua l i t y con
R E VtI Ero
W Sl

38. Krukenberg, K. A. et al. Conformational dynamics of Proc. Natl Acad. Sci. USA 101, 15005–15012 86. Wang, J. et al. Crystal structures of the HslVU
the molecular chaperone Hsp90. Q. Rev. Biophys. 44, (2004). peptidase–ATPase complex reveal an ATP-dependent
229–255 (2011). 63. Chaudhuri, T. K., Farr, G. W., Fenton, W. A., Rospert, S. proteolysis mechanism. Structure 9, 177–184 (2001).
39. Prodromou, C. et al. The ATPase cycle of Hsp90 drives & Horwich, A. L. GroEL/GroES-mediated folding of a 87. Effantin, G., Ishikawa, T., De Donatis, G. M.,
a molecular ‘clamp’ via transient dimerization of the protein too large to be encapsulated. Cell 107, Maurizi, M. R. & Steven, A. C. Local and global
N‑terminal domains. EMBO J. 19, 4383–4392 235–246 (2001). mobility in the ClpA AAA+ chaperone detected by
(2000). 64. Clare, D. K., Bakkes, P. J., van Heerikhuizen, H., cryo-electron microscopy: functional connotations.
40. Dutta, R. & Inouye, M. GHKL, an emergent ATPase/ van der Vies, S. M. & Saibil, H. R. Chaperonin complex Structure 18, 553–562 (2010).
kinase superfamily. Trends Biochem. Sci. 25, 24–28 with a newly folded protein encapsulated in the folding 88. Lander, G. C. et al. Complete subunit architecture of
(2000). chamber. Nature 457, 107–110 (2009). the proteasome regulatory particle. Nature 482,
41. Mickler, M., Hessling, M., Ratzke, C., Buchner, J. & 65. Clare, D. K. et al. ATP-triggered molecular mechanics 186–191 (2012).
Hugel, T. The large conformational changes of Hsp90 of the chaperonin GroEL. Cell 149, 113–123 (2012). 89. da Fonseca, P. C., He, J. & Morris, E. P. Molecular
are only weakly coupled to ATP hydrolysis. Nature Determines multiple conformations of ATP-bound model of the human 26S proteasome. Mol. Cell 46,
Struct. Mol. Biol. 16, 281–286 (2009). GroEL by single-particle cryo-EM and reveals a 54–66 (2012).
42. Siligardi, G. et al. Regulation of Hsp90 by the repertoire of domain movements that are likely to 90. Cranz-Mileva, S. et al. The flexible attachment of the
co‑chaperone Cdc37p/p50cdc37. J. Biol. Chem. 277, be involved in unfolding and encapsulation of N‑domains to the ClpA ring body allows their use on
20151–20159 (2002). non-native substrates. demand. J. Mol. Biol. 378, 412–424 (2008).
43. Prodromou, C. et al. Regulation of Hsp90 ATPase 66. Yifrach, O. & Horovitz, A. Nested cooperativity in the 91. Zhang, T. et al. Flexible connection of the N‑terminal
activity by tetratricopeptide repeat (TPR)-domain ATPase activity of the oligomeric chaperonin GroEL. domain in ClpB modulates substrate binding and the
co‑chaperones. EMBO J. 18, 754–762 (1999). Biochemistry 34, 5303–5308 (1995). aggregate reactivation efficiency. Proteins 80,
44. Chen, S. & Smith, D. F. Hop as an adaptor in the heat 67. Miyazaki, T. et al. GroEL–substrate–GroES ternary 2758–2768 (2012).
shock protein 70 (Hsp70) and Hsp90 chaperone complexes are an important transient intermediate of 92. Sauer, R. T. & Baker, T. A. AAA+ proteases: ATP-
machinery. J. Biol. Chem. 273, 35194–35200 the chaperonin cycle. J. Biol. Chem. 277, fueled machines of protein destruction. Annu. Rev.
(1998). 50621–50628 (2002). Biochem. 80, 587–612 (2011).
45. Retzlaff, M. et al. Asymmetric activation of the Hsp90 68. Nojima, T., Murayama, S., Yoshida, M. & Motojima, F. 93. Weber-Ban, E. U., Reid, B. G., Miranker, A. D. &
dimer by its cochaperone Aha1. Mol. Cell 37, Determination of the number of active GroES subunits Horwich, A. L. Global unfolding of a substrate protein
344–354 (2010). in the fused heptamer GroES required for interactions by the Hsp100 chaperone ClpA. Nature 401, 90–93
46. Theodoraki, M. A. & Caplan, A. J. Quality control and with GroEL. J. Biol. Chem. 283, 18385–18392 (1999).
fate determination of Hsp90 client proteins. Biochim. (2008). 94. Hinnerwisch, J., Fenton, W. A., Furtak, K. J.,
Biophys. Acta 1823, 683–688 (2012). 69. Buckle, A. M., Zahn, R. & Fersht, A. R. A structural Farr, G. W. & Horwich, A. L. Loops in the central
47. Southworth, D. R. & Agard, D. A. Client-loading model for GroEL–polypeptide recognition. Proc. Natl channel of ClpA chaperone mediate protein binding,
conformation of the Hsp90 molecular chaperone Acad. Sci. USA 94, 3571–3575 (1997). unfolding, and translocation. Cell 121, 1029–1041
revealed in the cryo‑EM structure of the human 70. Chen, L. & Sigler, P. B. The crystal structure of a (2005).
Hsp90:Hop complex. Mol. Cell 42, 771–781 (2011). GroEL/peptide complex: plasticity as a basis for 95. Glynn, S. E., Martin, A., Nager, A. R., Baker, T. A. &
Reveals a low-resolution cryo-EM structure of the substrate diversity. Cell 99, 757–768 (1999). Sauer, R. T. Structures of asymmetric ClpX hexamers
HSP90–HOP complex and suggests how HOP 71. Clare, D., Bakkes, P. J., van Heerikhuizen, H., reveal nucleotide-dependent motions in a AAA+
alters the conformation of HSP90 to one that is van der Vies, S. M. & Saibil, H. R. An expanded protein-unfolding machine. Cell 139, 744–756
poised to accept substrate and that closes upon protein folding cage in the GroEL–gp31 complex. (2009).
ATP binding. J. Mol. Biol. 358, 905–911 (2006). 96. Maillard, R. A. et al. ClpX(P) generates mechanical
48. Vaughan, C. K. et al. Structure of an Hsp90–Cdc37– 72. Chen, D.‑H. et al. Visualizing GroEL/ES in the act of force to unfold and translocate its protein substrates.
Cdk4 complex. Mol. Cell 23, 697–707 (2006). encapsulating a folding protein. Cell 153, Cell 145, 459–469 (2011).
49. Fang, L. Ricketson, D. Getubig, L. & Darimont, B. 1354–1365 (2013). Single-molecule force study suggests how the
Unliganded and hormone-bound glucocorticoid 73. Ranson, N. A. et al. Allosteric signaling of ATP ATP-dependent protease ClpXP mechanically
receptors interact with distinct hydrophobic sites in hydrolysis in GroEL–GroES complexes. Nature Struct. destabilizes and unfolds tagged GFP and
the Hsp90 C‑terminal domain. Proc. Natl Acad. Sci. Mol. Biol. 13, 147–152 (2006). translocates it through the pore channel in 1 nm
USA 103, 18487–18492 (2006). 74. Fei, X., Yang, D., LaRonde-LeBlanc, N. & Lorimer, G. H. steps.
50. Genest, O. et al. Uncovering a region of heat shock Crystal structure of a GroEL–ADP complex in the 97. DeLaBarre, B. & Brunger, A. T. Nucleotide dependent
protein 90 important for client binding in E. coli and relaxed allosteric state at 2.7 Å resolution. Proc. Natl motion and mechanism of action of p97/VCP. J. Mol.
chaperone function in yeast. Mol. Cell 49, 464–473 Acad. Sci. USA 110, E2958–E2966 (2013). Biol. 347, 437–452 (2005).
(2013). Determines the crystal structure of a highly 98. Niwa, H. et al. The role of the N‑domain in the ATPase
51. Trepel, J., Mollapour, M., Giaccone, G. & Neckers, L. asymmetric GroEL–ADP complex lacking key salt activity of the mammalian AAA ATPasep97/VCP.
Targeting the dynamic HSP90 complex in cancer. bridges. Shows that in this structure the J. Biol. Chem. 287, 8561–8570 (2012).
Nature Rev. Cancer 10, 537–549 (2010). conformations of individual subunits resemble the 99. Barthelme, D. & Sauer, R. T. Identification of the
52. Hemmingsen, S. M. et al. Homologous plant and range of states seen by cryo-EM. Cdc48•20S proteasome as an ancient AAA+
bacterial proteins chaperone oligomeric protein 75. Ditzel, L. et al. Crystal structure of the thermosome, proteolytic machine. Science 337, 843–846 (2012).
assembly. Nature 333, 330–334 (1988). the archaeal chaperonin and homolog of CCT. Cell 93, 100. Glover, J. R. & Lindquist, S. Hsp104, Hsp70, and
53. Goloubinoff, P., Gatenby, A. A. & Lorimer, G. H. GroE 125–138 (1998). Hsp40: a novel chaperone system that rescues
heat-shock proteins promote assembly of foreign 76. Kalisman, N., Adams, C. M. & Levitt, M. Subunit order previously aggregated proteins. Cell 94, 73–82
prokaryotic ribulose bisphosphate carboxylase of eukaryotic TRiC/CCT chaperonin by cross-linking, (1998).
oligomers in Escherichia coli. Nature 337, 44–47 mass spectrometry, and combinatorial homology 101. Goloubinoff, P., Mogk, A., Zvi, A. P., Tomoyasu, T. &
(1989). modeling. Proc. Natl Acad. Sci. USA 109, Bukau, B. Sequential mechanism of solubilization and
54. Ostermann, J., Horwich, A. L., Neupert, W. & 2884–2889 (2012). refolding of stable protein aggregates by a
Hartl, F. U. Protein folding in mitochondria requires 77. Leitner, A. et al. The molecular architecture of the bichaperone network. Proc. Natl Acad. Sci. USA 96,
complex formation with hsp60 and ATP hydrolysis. eukaryotic chaperonin TRiC/CCT. Structure 20, 13732–13737 (1999).
Nature 341, 125–130 (1989). 814–825 (2012). 102. Miot, M. et al. Species-specific collaboration of heat
55. Horwich, A. L. & Fenton, W. A. Chaperonin-mediated 78. Clare, D. K. et al. Multiple states of a nucleotide- shock proteins (Hsp) 70 and 100 in thermotolerance
protein folding: using a central cavity to kinetically bound group 2 chaperonin. Structure 16, 528–534 and protein disaggregation. Proc. Natl Acad. Sci. USA
assist polypeptide chain folding. Q. Rev. Biophys. 42, (2008). 108, 6915–6920 (2011).
83–116 (2009). 79. Zhang, J. et al. Mechanism of folding chamber closure 103. Lee, S. et al. The structure of ClpB: a molecular
56. Saibil, H. R., Fenton, W. A., Clare, D. K. & in a group II chaperonin. Nature 463, 379–383 chaperone that rescues proteins from an aggregated
Horwich, A. L. Structure and allostery of the (2010). state. Cell 115, 229–240 (2003).
chaperonin GroEL. J. Mol. Biol. 425, 1476–1487 80. Huo, Y. et al. Crystal structure of group II chaperonin 104. Weibezahn, J. et al. Thermotolerance requires
(2013). in the open state. Structure 18, 1270–1279 refolding of aggregated proteins by substrate
57. Braig, K. et al. The crystal structure of the bacterial (2010). translocation through the central pore of ClpB.
chaperonin GroEL at 2.8 Å. Nature 371, 578–586 81. Cong, Y. et al. Symmetry-free cryo‑EM structures of Cell 119, 653–665 (2004).
(1994). the chaperonin TRiC along its ATPase-driven 105. Seyffer, F. et al. Hsp70 proteins bind Hsp100
58. Xu, Z., Horwich, A. L. & Sigler, P. B. The crystal conformational cycle. EMBO J. 31, 720–730 regulatory M domains to activate AAA+ disaggregase
structure of the asymmetric GroEL–GroES–(ADP)7 (2011). at aggregate surfaces. Nature Struct. Mol. Biol. 19,
chaperonin complex. Nature 388, 741–750 (1997). 82. Horovitz, A. & Willison, K. R. Allosteric regulation 1347–1355 (2012).
59. Farr, G. W. et al. Multivalent binding of nonnative of chaperonins. Curr. Opin. 15, 646–651 106. Rosenzweig, R., Moradi, S., Zarrine-Afsar, A., Glover, J.
substrate proteins by the chaperonin GroEL. Cell 100, (2005). R. & Kay, L. E. Unraveling the mechanism of protein
561–573 (2000). 83. Bigotti, M. G. & Clarke, A. R. Cooperativity in disaggregation through a ClpB–DnaK interaction.
60. Elad, N. et al. Topologies of a substrate protein bound the thermosome. J. Mol. Biol. 348, 13–26 (2005). Science 339, 1080–1083 (2013).
to the chaperonin GroEL. Mol. Cell 26, 415–426 84. Neuwald, A. F., Aravind, L., Spouge, J. L. & Proposes a model based on data from methyl
(2007). Koonin, E. V. AAA+: a class of chaperone-like ATPases TROSY NMR analysis in which the DnaK ATPase
61. Lin, Z., Madan, D. & Rye, H. S. GroEL stimulates associated with the assembly, operation, and domain binds the coiled-coil domain of ClpB to
protein folding through forced unfolding. Nature disassembly of protein complexes. Genome Res. 9, activate disaggregation in this bi‑chaperone system.
Struct. Mol. Biol.15, 303–311 (2008). 27–43 (1999). 107. Wendler, P. et al. Atypical AAA+ subunit packing
62. Motojima, F., Chaudhry, C., Fenton, W. A., Farr, G. W. 85. Hanson, P. I. & Whiteheart, S. W. AAA+ proteins: creates an expanded cavity for disaggregation by the
& Horwich, A. L. Substrate polypeptide presents a have engine, will work. Nature Rev. Mol. Cell Biol. 6, protein-remodeling factor Hsp104. Cell 131,
load on the apical domains of the chaperonin GroEL. 519–529 (2005). 1366–1377 (2007).

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 14 | O CTOBER 2013 | 641

© 2013 Macmillan Publishers Limited. All rights reserved

REVIEWS

108. Lee, S., Sielaff, B., Lee, J. & Tsai, F. T. CryoEM structure 114. Buxbaum, J. N. & Linke, R. P. A molecular history of 122. Shiau, A. K., Harris, S. F., Southworth, D. R. &
of Hsp104 and its mechanistic implication for protein the amyloidoses. J. Mol. Biol. 421, 142–159 (2012). Agard, D. A. Structural analysis of E. coli hsp90
disaggregation. Proc. Natl Acad. Sci. USA 107, 115. Eisenberg, D. & Jucker, M. The amyloid state of reveals dramatic nucleotide-dependent conformational
8135–8140 (2010). proteins in human diseases. Cell 148, 1188–1203 rearrangements, Cell 127, 329–340 (2006).
109. Wang, F. et al. Structure and mechanism of the (2012). 123. Dollins, D. E., Warren, J. J., Immormino, R. M. &
hexameric MecA-ClpC molecular machine. Nature 116. Auluck, P. K., Chan, H. Y. E., Trojanowski, J. Q., Gewirth, D. T. Structures of GRP94–nucleotide
471, 331–335 (2011). Lee, V. M. Y. & Bonini, N. M. Chaperone suppression complexes reveal mechanistic differences between the
110. Oguchi, Y. et al. A tightly regulated molecular toggle of α‑synuclein toxicity in a Drosophila model for hsp90 chaperones. Mol. Cell 28, 41–56 (2007).
controls AAA+ disaggregase. Nature Struct. Mol. Parkinson’s disease. Science 295, 865 − 868 (2002). 124. Sousa, M. C. et al. Crystal and solution structures of
Biol. 19, 1338–1346 (2012). 117. Kitamura, A. et al. Cytosolic chaperonin prevents an HslUV protease–chaperone complex. Cell 103,
Uses mutations at opposite ends of the ClpB polyglutamine toxicity with altering the aggregation 633–643 (2000).
coiled-coil domain together with accessibility and state. Nature Cell Biol. 8, 1163–1170 (2006).
interaction studies and reveals a toggle switch 118. Tam, S., Geller, R., Spiess, C. & Frydman, J. The Acknowledgements
mechanism for regulating disaggregation activity. chaperonin TRiC controls polyglutamine aggregation The author is grateful to D. Clare, C. Vaughan, J. Trapani and
111. Kaushal, S. & Khorana, H. G. Structure and function in and toxicity through subunit-specific interactions. D. Middendorf for helpful comments on the manuscript and
rhodopsin. 7. Point mutations associated with Nature Cell Biol. 8, 1155–1162 (2006). thanks the Wellcome Trust for funding.
autosomal dominant retinitis 119. Behrends, C. et al. Chaperonin TRiC promotes the
pigmentosa. Biochemistry 33, 6121–6128 assembly of polyQ expansion proteins into nontoxic Competing interests statement
(1994). oligomers. Mol. Cell 23, 887 –897 (2006). The author declares no competing financial interests.
112. Mendes, H. F., Van der Spuy, J., Chapple, J. P. & 120. Balch, W. E., Morimoto, R. I., Dillin, A. & Kelly, J. W.
Cheetham, M. E. Mechanisms of cell death in Adapting proteostasis for disease intervention.
rhodopsin retinitis pigmentosa: implications for Science 319, 916–919 (2008). SUPPLEMENTARY INFORMATION
therapy. Trends Mol. Med. 11, 177–185 (2005). 121. Powers, E. T., Morimoto, R. I., Dillin, A., Kelly, J. W. & See online article: S1 (table) | S2 (movie) | S3 (movie) |
113. Chiti, F. & Dobson, C. M. Protein misfolding, functional Balch, W. E. Biological and chemical approaches to S4 (figure)
amyloid, and human disease. Annu. Rev. Biochem. 75, diseases of proteostasis deficiency. Annu. Rev. ALL LINKS ARE ACTIVE IN THE ONLINE PDF
333–366 (2006). Biochem. 78, 959–991 (2009).

642 | O CTOBER 2013 | VOLUME 14 www.nature.com/reviews/molcellbio

© 2013 Macmillan Publishers Limited. All rights reserved