Love it or cleave it: Tough choices in protein quality control
Michael R. Maurizi
Laboratory of Cell Biology, National Cancer
Institute. Bethesda, Maryland 20892,
USA.mmaurizi@helix.nih.gov
Housekeeping is not a humdrum process when it comes to protein
quality control. Inside chaperone-protease machines, proteins are subject to
dramatic life or death trials and must shape up rapidly to survive.
Accumulation of aberrantly folded proteins is the basis of many diseases, so
the outcome is vitally important to the cell.
Chaperone-protease machines have assumed a major position in the
hierarchy of important biological molecules that are of interest to cell
biologists, biochemists, and structural biologists. These complex proteins are
associated with many fundamental processes required for the life of a cell. In
addition, the molecules themselves have fascinating and intricate structural
designs and mechanisms of action. Until recently the ATP-dependent molecular
chaperone-protease machines, such as the proteasomes, the Clp proteases, Lon
proteases and the membrane-associated FtsH/Yme1 proteases have received the
most attention. Now, the crystal structures of two members of the DegP/HtrA
protein family — the mammalian HtrA2/Omi, reported by Li et
al.1 on page 436 of this issue
of Nature Structural Biology and the Escherichia coli DegP/HtrA,
reported by Krojer et al.2 in a recent issue of
Nature — have broadened the focus to include a novel class of
ATP-independent chaperone-proteases.
DegP is an essential protein in the E.coli periplasm. It protects
cells by degrading damaged or denatured proteins generated during heat or
chemical stress3. Such damaged proteins would otherwise form
aggregates that are highly toxic to cells. DegP homologs in a number of
pathogenic bacteria have been shown to be virulence factors, helping bacteria
survive within the host cell in part by degrading periplasmic proteins damaged
by the host's oxidative defense system4. Several years ago,
Erhmann and colleagues5 reported an unusual property of E.
coli DegP. At temperatures below 28 °C, DegP acts solely as a
chaperone, protecting unfolded proteins from irreversible aggregation and
increasing the yield of refolded protein. Above 28 °C, its protease
activity increases dramatically and DegP efficiently degrades the unfolded
proteins. Thus, DegP exhibits dual function and appears to combine chaperone
activity with protease activity. While precedents exist for chaperone activity
contributing to protein degradation, an important question is how this coupling
is accomplished without the expenditure of energy. Equally puzzling are how
proteolytic activity is suppressed at low temperature and what the nature of
the molecular switch is that favors the degradative over the refolding
pathway.
Interest in HtrA proteases was further stimulated by the discovery of
mammalian homologs and their association with cellular stress responses6. Human HtrA1 is elevated in osteoarthritic cartilage7,
whereas HtrA2 interacts with various IAPs (inhibitor of apoptosis proteins)
and, when over-expressed, causes cell death8. The activity of
XIAP, a mammalian caspase-9 inhibitor, is antagonized by, among other things,
proteins such as Smac/DIABLO9, which is released from
mitochondria in response to certain cell stress signals. Both Smac/DIABLO and
HtrA2 specifically interact with IAPs through similar tetrapeptide motifs near
their N-termini. HtrA2 must be proteolytically active to exert its killing
effect, and, unlike the bacterial enzyme, it has proteolytic activity at both
low and high temperatures. How HtrA2 proteolytic activity is regulated is not
known.
HtrA subunits comprise a proteolytic domain and either one (HtrA2) or
two (DegP) peptide-binding (PDZ) domains at the C-terminus. The crystal
structures suggest that activities of HtrA2 and DegP derive from the interplay
between two different structural features—the local environment of the
proteolytic active site and the positions of the PDZ domains. Both proteases
form trimers connected through their proteolytic domains. Trimer contacts are
made through a narrow region near the N-terminus giving rise to a hollow
pyramid-shaped particle with the active sites facing inward. Access to the
active sites is restricted, identifying the HtrA family as a novel example of
self-compartmentalized proteases10. In HtrA2, access is
restricted by the PDZ domains, which rotate inward and form the base of the
pyramid. In DegP, the native structure is actually a dimer of trimers joined by
intertwined loops extending out from the base of the pyramid, thus forming a
bipyramid with the active sites completely sequestered within (Fig. 1).
Figure 1. HtrA proteases from E. coli and from mammals use
peptide-binding PDZ domains to screen appropriate substrates and to sweep
unfolded proteins into the sequestered proteolytic chamber.
Mammalian HtrA2/Omi (top) has a single PDZ domain per
subunit. In the oligomer, the PDZ domains block the active site until a
substrate is encountered. Are activating receptors also used to facilitate the
process by displacing the PDZ domains and exposing access channels to the
active site chamber? In E. coli DegP/HtrA (bottom), tandem PDZ domains
in each subunit are used to first bind substrates tightly at multiple sites.
Rapid reciprocating movement of tentacular arms of PDZ domains would put stress
on the protein and cause it to unfold. The home position of the PDZ domains
aligns the peptide-binding grooves with the access channels and allows facile
passage of the extended polypeptide into the interior chamber. There the
protein is degraded, but when the protease is in an inactive state at low
temperature, the unfolded protein is extruded instead and can refold. As a
chaperone, DegP acts slowly on stoichiometric amounts of substrate and
presumably uses thermal energy to tease the protein incrementally into a
folding-competent form.
Another striking difference in the structures of HtrA2 and DegP is in
the environments of their active sites. The proteolytic domains are typical of
trypsin-like proteases with a classical Asp-His-Ser catalytic triad located in
a crevice between two -barrel lobes. In HtrA2, the proteolytic active
sites are in an active conformation poised for catalysis. In contrast, the
active site of DegP is sterically blocked by a trio of loops from the
surrounding protein, and the catalytic triad is distorted away from an active
configuration. The major blocking loop is contributed by the symmetry-related
subunit in the adjacent pyramid, suggesting that interaction of two trimers may
be a significant factor in formation of the proteolytically inactive state of
DegP.
A surprising aspect of the two structures is the apparent contradictory
roles played by the PDZ domains. They appear to be the major element hindering
access to the proteolytic active sites in HtrA2 but in DegP are poised to
facilitate entry of substrates into the active site chamber (Fig. 1). PDZ domains are widespread and highly conserved
protein interaction domains with a unique peptide-binding motif (GLGF or its
conservative variations) that interacts with extended conformations of exposed
C-terminal regions of proteins11. The domain contains
predominantly -strands. The bound peptide is insinuated between a short
helix and the edge of the adjacent -strand, completing the -sheet
and adding to its stability. PDZ domains generally participate in assembly and
organization of protein complexes, although in at least one other case, the
bacterial tail-specific protease (Tsp) and its eukaryotic homolog, CtpA12, these domains are used to promote association between a protease
and targeted protein substrates carrying C-terminal degradation tags.
Several lines of evidence suggest that the PDZ domains of HtrA2 play a
dual role in its function. Because the active sites are sequestered in part by
the protein interaction domain, it was reasonable to expect that removing the
PDZ domain might allow unregulated proteolytic activity. Li et al.1 generated a mutant HtrA2 lacking the PDZ domain. This mutant has very
high protease activity, and moreover is efficient at cell killing. Notably,
monomeric forms of HtrA2 are proteolytically inactive, but removal of their PDZ
domains restores the activity. On the basis of the requirement for a trimer and
the need to displace the PDZ domain to activate protease activity, Li et
al.1 speculated that trimeric HtrA2 might interact through
its PDZ domains with a trimeric ligand, such as a cell death receptor, which
displaces the PDZ domain and makes the active site chamber accessible to
substrates.
What adds to the attractiveness of this model is the structure of DegP,
which at first glance appears very different from HtrA2. Two different hexamers
were found in the crystal of DegP, one in an 'open' and one in a 'closed'
conformation. Both molecules are bipyramids connected through relatively narrow
extended loops that form pillars separating the bases (Fig.
1). The access routes to the interior chamber lie between the pillars
and can accommodate little more than secondary structure elements. The two PDZ
domains per subunit do not impinge on the active site cleft but rather are
splayed out from the base of the pyramid in each half of the particle and form
mobile 'side walls'. In the closed conformation, interactions between pairs of
the 12 PDZ domains completely block the entrances to the interior chamber. In
the open conformation, the PDZ1 domains from opposite pyramids do not interact
and their peptide binding grooves are positioned to shunt extended segments of
substrates through the side openings into the active site chamber (Fig. 1).
The position of PDZ1 of DegP could well be a model for the position of
the PDZ domain of HtrA2 upon activation by trimeric receptors. Interaction
might both relieve inhibition and position the PDZ domains to facilitate the
passage of substrates into the degradative chamber. Li et al.1 proposed that the activating receptor interacts with HtrA2 through
the PDZ domain, which could put the receptor in competition with the substrate.
An alternative model might be that receptor interaction occurs through another
site in HtrA and, through an allosteric mechanism, causes the PDZ domains to
adopt an active orientation, thus leaving the peptide-binding groove free for
interaction with substrate. In any event, the mobility of the PDZ domains
implied in the structure of DegP suggests that a domain-displacement model for
activation of HtrA2 is quite feasible.
The mammalian enzyme does not appear to have chaperone activity
independent of proteolytic activity, even at low temperature. The primary
explanation for this appears to be the integrity of the proteolytic site, which
is not blocked by a loop from another interacting trimer, as observed in the
structure of DegP. In mammalian HtrA proteins, the major 'blocking loop' is
shorter than that of the E. coli enzyme by 19 amino acids, thereby
precluding its serving as a dimerization interface. Possibly, the shorter loop
in the mammalian protein, which shares a number of conserved residues with its
bacterial homolog, can have a regulatory effect within the trimer or interact
with another protein.
The E. coli DegP appears to have three conformations: closed,
open chaperone, and open protease. Because the proteolytic domains are
virtually identical in both states, the transition between open and closed
forms demonstrates the considerable flexibility of the hinge connecting the
proteolytic and PDZ1 domains and the mobility of the PDZ2 domain. The function
of the closed molecule, which appears inert because most of the binding sites
are not accessible, is not known, but it is not believed to be the chaperone
form. The structure of the open form in the crystal, with its non-functional
proteolytic site, should represent the chaperone state. What conditions in
addition to changes in temperature convert the protein between these states is
of great interest. Krojer et al.2 noted that the conserved
Arg 262 residue, which is positioned to interact with bound substrate, is
located in the hinge between the protease and PDZ domains. Substrate binding
could trigger movement of the PDZ domain and help favor the transition between
states. Since, along with the carboxylate binding site, the -2 and
-3 binding pockets in the PDZ domain dictate binding specificity,
different substrates could impose this conformation switch with different
efficiencies. Presumably a change in the trimer interface could alter the
position of the major blocking loop, allowing reconfiguration of the catalytic
triad.
Several aspects of the structures remain to be explained, and in
particular the mechanism of coupling between peptide binding to the PDZ domain
and either proteolytic activity or conformational switching remains to be
elucidated. In HtrA2, a potentially inhibitory interaction is produced by the
juxtapositioning of the loop between 5 and 6 with the peptide
binding groove in the PDZ domain, which could exclude binding of substrates to
that site. This interaction might stabilize an inactive conformation of the
enzyme or could be involved in modulating substrate binding or release from the
PDZ binding groove. Since the residues involved in binding are near the active
site, their displacement by substrate could also have an activating effect on
the protease. In E. coli DegP, a comparable function is proposed for a
loop at the beginning of PDZ2, which interacts with its carboxylate binding
groove. This interaction is postulated to facilitate displacement of substrate
and to play a role in substrate translocation.
HtrA proteases act primarily on unfolded or misfolded proteins. The
binding sites for the exposed hydrophobic regions of unfolded proteins are
within the chamber, but the geometric constraints allow binding of an extended
conformation only. The binding pockets of the PDZ domains are also hydrophobic,
although they have greater specificity requirements and are generally more
selective. Substrates must encounter the more specific exterior sites before
gaining access to the less selective interior chamber. This situation is not
unlike that in the Clp proteases or the proteasome, in which large ATPase
complexes bind to the face of the proteolytic component and block random access
to the entry channel, while at the same time providing binding sites for
specific substrates and a means of actively translocating them inside. Kroger
et al.2 speculate that the local and global fluidity of
the PDZ domains, which results in continuous formation and dissolution of
contacts with exposed regions of the unfolded protein substrates of DegP, would
allow the protease to capture substrates and deliver them to the entrances to
the inner cavity. From there the protein would be subject to further unfolding
and either degradation or, for DegP at low temperature, release in a state
favoring refolding. Now that the structural elements are in place, the exciting
process of teasing the intricate processes of discrimination and decision
making in the activity of these important and potentially lethal machines can
begin in earnest.
-barrel lobes. In HtrA2, the proteolytic active
sites are in an active conformation poised for catalysis. In contrast, the
active site of DegP is sterically blocked by a trio of loops from the
surrounding protein, and the catalytic triad is distorted away from an active
configuration. The major blocking loop is contributed by the symmetry-related
subunit in the adjacent pyramid, suggesting that interaction of two trimers may
be a significant factor in formation of the proteolytically inactive state of
DegP.
