Chapter 61

Molecular Chaperone Proteins

MARK MAYHEW and FRANZ-ULRICH HARTL

INTRODUCTION

How newly synthesized polypeptide chains fold into a complex but well-defined three-dimensional structure is of fundamental importance in biology. Remarkably, only recently has the process of protein folding in vivo become the focus of intensive research, stimulated by the discovery of a cellular protein machinery that catalyzes it. The components of this machinery are the so-called molecular chaperones and certain folding catalysts, such as protein disulfide isomerases and prolyl isomerases (31, 40, 60, 110). The molecular chaperones are the focus of this chapter. Before describing their functions, however, it is useful to discuss briefly some of the basic principles of protein folding.

Traditionally, the protein-folding problem has been studied in vitro with purified, chemically denatured polypeptides as substrates. The foundation of this active field was laid by Anfinsen and colleagues more than three decades ago in experiments demonstrating that correct (re)folding of small proteins, such as RNase, was possible upon removal of denaturant (2). This led to the conclusion that all of the information required for the folding of a protein resides in its primary sequence of amino acid residues and that proteins are capable of reaching their native state, presumably their most stable conformation, in a spontaneous reaction, essentially without the participation of additional factors. The wealth of in vitro studies that followed these key experiments has indeed shaped our understanding of the general kinetic and thermodynamic principles of protein folding, at least for small, globular proteins (Fig. 1) (5, 105, 121). Briefly, within a few milliseconds of the initiation of folding, the extended polypeptide chain collapses, driven primarily by hydrophobic forces, to a compact folding intermediate. This intermediate, aptly named the molten globule, is stabilized by still flexible interactions between hydrophobic residues in the core of the molecule and possesses elements of secondary structure, such as α-helices and β-sheets, but lacks the ordered and stable interaction between these elements which form the tertiary structure in the native state (19, 21, 120, 122, 123). Thus, the hydrophobic core of the molten globule is not yet stable. This leads to the exposure of hydrophobic protein surfaces to solvent, which makes this intermediate highly prone to aggregation, an off-pathway folding reaction that is considered irreversible under most conditions (Fig. 1). As a consequence, in vitro protein-folding studies are usually performed with low concentrations of protein and below physiological temperatures. Folding from the "molten-globule" intermediate toward the native state is a comparatively slow process that occurs on the time scale of seconds to minutes (Fig. 1). As the polypeptide progresses along this pathway, it is believed to pass through various incompletely folded intermediates by a series of structural rearrangements. Some of these intermediates, although being thermodynamically less stable than the native state, may be separated from each other by considerable barriers of activation energy. In some cases, these energy barriers are so high that the apparent end point of the folding reaction is determined kinetically rather than thermodynamically (4).

Fig. 1In vitro protein-folding pathway. An unfolded polypeptide in an aqueous environment will very rapidly (within milliseconds) collapse into a compact intermediate that is often referred to as a molten globule (see the text). This intermediate is prone to aggregation because of the exposure of hydrophobic surfaces. Aggregation is favored by conditions such as elevated temperature and high protein concentration. In a suitable environment, the compact intermediate will progress on the folding pathway toward the native state, possibly via a number of intermediates.

Source:

The conclusion from the Anfinsen experiment that the polypeptide chain contains the information required to specify the conformation of the native state still holds true today. However, the concept of spontaneous protein folding is an oversimplification, particularly when considering protein folding in the cell. Many proteins built of multiple domains (folding units of 50 to 200 amino acid residues) will not efficiently refold in vitro and frequently end up as insoluble aggregates. This may be due at least in part to incorrect interactions between different parts of the polypeptide at the stage of rapid and indiscriminate chain collapse that marks the beginning of in vitro refolding experiments. Depending on the concentration of such folding intermediates, aggregation would be more or less rapid. In the cell, a nascent polypeptide emerging from the ribosome would be in an extremely dense environment with a total concentration of folding chains as high as 30 to 50 μM in the Escherichia coli cytosol, conditions that should strongly favor aggregation over folding. In principle, the sequential extension of the polypeptide during translation may have the advantage of allowing the separate folding of protein domains. However, at the same time it aggravates the danger of misfolding and aggregation, because the formation of tertiary structure is known to be a highly cooperative process that requires at least the complete sequence of an independent protein domain to be productive. Until such a domain has emerged from the ribosome, the nascent chain will at best be able to adopt the conformation of a partially folded intermediate, a state which is very prone to aggregation. Folding will not occur until a suitable chain length is achieved, a process that may take minutes. How, then, is the growing nascent chain stabilized against misfolding and aggregation until the point is reached at which folding can be productive?

The solution to this question has emerged with the identification and characterization of a large number of proteins collectively named molecular chaperones. These proteins, which can be subdivided into structurally distinct families, have the general ability to bind to unfolded and nascent polypeptide chains, thus preventing their aggregation. Subsequently, the polypeptide substrate is released in a controlled mechanism, often regulated by ATP binding and hydrolysis, that may lead to correct folding. By definition, molecular chaperones bind a broad spectrum of unfolded or partially folded proteins, apparently by interacting with their exposed hydrophobic residues and surfaces. However, chaperones never form part of the final native protein or assembled complex. Interestingly, a number of molecular chaperones are also heat shock proteins (hsps), that is, their expression is inducible by a variety of cellular stresses including elevated growth temperatures (see chapter 88). It is now clear, however, that these proteins also play essential roles under nonstress conditions in protein folding and related processes.

As described in the following sections, at least five structurally distinct families of molecular chaperones have been characterized in E. coli (Table 1). Members of these classes may cooperate functionally in various protein-folding, assembly, and disassembly processes (22, 36, 39, 47, 52, 58, 68, 127). DnaK, a member of the hsp70 family, functions together with the chaperone DnaJ, an hsp40 family member, in stabilizing nascent polypeptides on ribosomes, in phage λ DNA replication and similar processes, and in the regulation of the heat shock response. A third type of chaperone, GroEL, is a member of the hsp60 family. Together with the cofactor GroES, GroEL mediates the productive folding of newly synthesized polypeptides to their native conformations. GroEL/GroES may cooperate in this process with DnaK and DnaJ. A fourth type of chaperone, SecB, is thought to specifically bind proteins destined for export, stabilizing them in an unfolded state competent for membrane translocation. Finally, PapD, a highly specialized chaperone, differs considerably from the proteins mentioned above. It resides in the periplasmic space and interacts with pilus subunit proteins after they have been exported from the cytosol.

Table 1Molecular chaperones of E. coli

CYTOSOLIC PROTEIN FOLDING

DnaK, the hsp70 Family Member of E. coli

Members of the hsp70 family of molecular chaperones have the capability to bind to a wide range of nonnative polypeptide substrates. In eukaryotic cells, these proteins were originally identified by virtue of their induction under cellular stress (144). The main hsp70 homolog of E. coli, DnaK, was first identified in mutant strains defective in phage λ replication. It is both constitutively expressed and inducible by heat shock or metabolic stress (also see chapter 88 and reference 37 for a review).

Evidence for a role of hsp70 in binding nonnative proteins was first obtained from studies of eukaryotic family members. The hsp70 homolog in the endoplasmic reticulum of mammalian cells was found associated with immunoglobulin heavy chains that had not yet assembled with light chains (48, 108) and was therefore given the name BiP for immunoglobulin heavy- chain binding protein. It was further observed that in response to heat shock in mammalian cells, hsp70 accumulated in the nucleolus, apparently bound to partially assembled ribosomes (158). Release of hsp70 was thought to require ATP hydrolysis (92). These observations led to the proposal that during heat shock, proteins become destabilized and expose hydrophobic regions, resulting in misfolding and aggregation. By binding to these regions, hsp70 may prevent and perhaps even reverse aggregation (118). Although it now seems clear that ATP hydrolysis by hsp70 is not required for the release step (see below), important aspects of these predictions for the possible function of hsp70s have since been substantiated experimentally.

Evidence for a potential role of hsp70 in facilitating the folding of nascent polypeptide chains as they emerge from the ribosome was first presented by Beckmann et al. (8). When mammalian cells were radiolabeled during protein synthesis and lysed and the lysates were depleted of ATP to prevent substrate release from hsp70, immunoprecipitation with hsp70 antibody resulted in the coprecipitation of many newly synthesized polypeptide chains ranging in size from 20 to 200 kDa. This suggested that hsp70 bound to nascent polypeptide chains as they were being synthesized on the ribosome (8). More recent studies have shown that the interaction of hsp70 with translating ribosomes is disrupted by the addition of puromycin, a reagent known to release nascent chains from ribosomes (111), and that cotranslational binding of hsp70 to nascent polypeptide chains is a prerequisite for their correct folding (32). Furthermore, DnaK has been demonstrated to cofractionate with polyribosomes and to interact with nascent polypeptides (34, 78). Together, these studies clearly demonstrate the importance of the association of hsp70 with nascent chains and newly synthesized polypeptides and implicate DnaK as an important factor in in vivo protein folding in the E. coli cytosol.

Nucleotide and Peptide Binding by hsp70

All members of the hsp70 family of molecular chaperones are composed of a highly conserved 44-kDa amino-terminal ATPase domain and a ∼25-kDa carboxy-terminal domain that contains the binding site for unfolded polypeptide. The structure of the ATPase domain has been solved for bovine hsc70, a constitutively expressed member of the hsp70 family (29). The nucleotide- binding core of this domain bears a striking resemblance to that of hexokinase and G-actin, both of which are known to undergo dramatic conformational changes on ATP binding and release. This suggests that hsc70 uses a similar phosphotransferase mechanism and undergoes similar nucleotide-induced conformational changes (29). Indeed, different patterns of proteolytic fragments obtained by trypsin digestion of DnaK in the presence and absence of nucleotide have been interpreted as reflecting such conformational changes (96).

The three-dimensional structure of the C-terminal substrate-binding domain of hsp70 is still unclear (15, 155). A consensus secondary structure prediction of this domain, based on 33 aligned hsp70 proteins, could be tentatively superimposed onto the secondary structure of the peptide-binding domain of the major histocompatibility complex class I molecule (128). However, the significance of this prediction has recently been questioned as a result of preliminary structural studies of the C-terminal domain of hsc70 obtained by nuclear magnetic resonance (NMR) spectroscopy (109). Nevertheless, as in the case of the major histocompatibility complex class I molecule, hsp70 appears to bind short peptides by virtue of hydrophobic interactions. The specificity of substrate binding to the endoplasmic reticulum hsp70, BiP, has been investigated by measuring the ability of random peptides to stimulate its ATPase activity (30). The peptides that bound well to BiP were sequenced and shown to be predominantly hydrophobic and optimally heptameric in length (30). More recently, libraries of bacteriophages displaying random octa- or dodecapeptides were screened to characterize those peptides that bound to BiP (9). This approach demonstrated that the substrate peptides that were recognized by BiP contain aromatic or hydrophobic amino acids at alternate positions, suggesting that the peptide binds in an extended conformation. A model based on these results proposed that the substrate-binding site of BiP contains four pockets that can each accommodate the side chain of a hydrophobic or aromatic residue and that occupancy of at least two of these sites is sufficient for stable binding to BiP (Fig. 2) (9). While the subtleties of the two studies are different, the general conclusion is that hsp70 recognizes hydrophobic patches on nascent, transiently unfolded, or destabilized proteins. The conformation of a substrate when bound to hsp70 was recently examined directly by analyzing transferred nuclear Overhauser effects in two-dimensional NMR spectroscopy (86). This study demonstrated that a 13-residue peptide bound to DnaK in an extended conformation, consistent with the role of DnaK in binding to nascent chains.

Fig. 2Peptide binding to the hsp70 homolog BiP. The substrate polypeptide is shown as an extended chain from which the side chains of residues 2, 4, 6, and 8 would fit into four pockets in the peptide-binding site of BiP. Generally, the specificity at each of the four sites favors large hydrophobic or aromatic residues; however, some variation is seen, as detailed in the figure.

Source: Adapted from reference 9.

The ATPase Cycle of DnaK, DnaJ, and GrpE

How exactly nucleotide-dependent conformational changes in the ATPase domain are transferred to the carboxy-terminal domain of hsp70 is not known. Recently, Palleros et al. demonstrated that it is only the binding of ATP and not its hydrolysis that is required for polypeptide dissociation from DnaK (115). The peptide affinity of hsp70 is higher in the ADP-bound state than in the ATP-bound state (116). On the other hand, peptide binding and peptide release are most rapid in the ATP state. It has been proposed, therefore, that upon peptide binding, DnaK hydrolyzes its bound ATP, resulting in the ADP state that binds peptide more tightly. This interconversion of the two states of DnaK is modulated by two protein cofactors, DnaJ and GrpE. DnaJ belongs to the class of hsp40 proteins and has itself been described as a molecular chaperone, on the basis of its ability to bind nonnative polypeptides and prevent protein aggregation (89). GrpE is a ∼20-kDa heat shock protein without chaperone function. It acts merely as a nucleotide exchange factor for DnaK (95). DnaK, DnaJ, and GrpE cooperate in a reaction cycle that has been studied in an in vitro refolding reaction with firefly luciferase as the substrate. The current model of this reaction cycle can be broken down into the following defined steps. (i) DnaJ binds the unfolded polypeptide. (ii) DnaJ interacts with DnaK, presenting the polypeptide to ATP-bound DnaK that has a fast on-rate for peptide association. (iii) DnaJ activates the hydrolysis of ATP by DnaK and stabilizes the ADP state of DnaK that holds the polypeptide in a tightly bound state. One round of ATP hydrolysis is necessary to form a stable ternary complex of substrate polypeptide, DnaK, and DnaJ. (iv) This complex is resolved by the action of GrpE, which catalyzes the dissociation of ADP from DnaK, thus labilizing its interaction with DnaJ and polypeptide. (v) ATP binding then releases the bound polypeptide. At this point, the polypeptide may fold, be transferred to another chaperone, or rebind to DnaJ and DnaK (Fig. 3) (143). The combined action of DnaJ and GrpE leads to a marked stimulation of the otherwise very slow ATPase activity of DnaK (95). Cycles of chaperone binding and release may serve to stabilize incomplete polypeptides during translation and allow folding once synthesis of the protein (or a protein domain) has been completed.

Fig. 3Model for the reaction cycle of DnaK, DnaJ, and GrpE in protein folding. 1. DnaJ binds to unfolded (U) or partially folded (I) polypeptide. 2. An interaction of DnaJ with DnaK in its ATP-bound state presents the polypeptide for binding to DnaK. 3. DnaJ induces ATP hydrolysis in DnaK, which results in the stabilization of the DnaK-polypeptide interaction. 4. GrpE induces the release of ADP from DnaK, which in turn weakens the ternary complex. 5. The rebinding of ATP to DnaK causes the release of the bound polypeptide, which may fold to the native state (N), rebind to DnaJ, or be bound by another chaperone protein such as GroEL.

Source:

Functional Cooperation of DnaK with DnaJ and GrpE

DnaK, DnaJ, and GrpE also cooperate in reactions other than protein folding to dissociate specific protein-protein interactions, thus regulating the functional activity of these components. Examples of such activities are seen in the replication of bacteriophage λ and of plasmids mini-P1 and mini-F (37, 98). The initiation of phage λ replication depends on the recruitment of the DnaB helicase to the origin of replication. However, DnaB is bound into a complex with the λ O and λ P proteins (the preprimosomal complex), in which the helicase is inactive. Activation of DnaB requires the disassembly of this complex to begin unwinding the DNA, a prerequisite for replication. This task is performed by DnaK in cooperation with DnaJ and GrpE. In vitro disassembly reactions, in which DnaK and DnaJ bind to the preprimosomal complex and release the λ P protein in an ATP- and GrpE-dependent process, have been characterized (1, 94, 96, 163). Significantly, the ATP-dependent formation of a ternary complex between the λ P protein, DnaK, and DnaJ has been described as an essential step in the reaction (61). This complex dissociates upon addition of GrpE.

In phage λ replication and several similar processes, DnaK and DnaJ act on specific proteins which are apparently fully folded. This constitutes an important difference to the more general functions of the DnaK, DnaJ, and GrpE chaperone team in interacting with a wide variety of nonnative polypeptide substrates, preventing the aggregation of polypeptide chains during folding, membrane translocation, and degradation. Nevertheless, the basic principle of action to modulate protein conformation from nonnative to native or from inactive to active seems to be preserved in all these cases. The three proteins may even have a certain capability of dissociating the nonspecific aggregates of proteins that may form under heat stress. In vivo evidence supporting this view was provided by Gaitanaris et al., who demonstrated that heat-denatured λ repressor could be renatured by DnaK, DnaJ, and GrpE in intact E. coli cells (33). A second study, in vitro, demonstrated that DnaK could protect RNA polymerase from heat inactivation and, at high molar excess, rescue heat-inactivated RNA polymerase in an ATP-dependent manner (137).

More recently, Schröder et al. provided interesting mechanistic insight by studying the DnaK-, DnaJ-, and GrpE-dependent reactivation of the thermolabile protein firefly luciferase both in E. coli and in vitro (135). Luciferase-expressing cells were exposed to heat shock at 42°C and then allowed to recover under conditions in which further protein synthesis was inhibited. Under these conditions, luciferase regained 50% of its initial activity. When this experiment was performed with cells carrying mutations in either the dnaK or dnaJ gene, reactivation was not observed, and in cells carrying a mutation in the grpE gene, reactivation was at least reduced (135). In vitro experiments with pure chaperone proteins and luciferase demonstrated that while the chaperones did not prevent the inactivation of luciferase, they were indeed essential for its ATP-dependent reactivation. Interestingly, the chaperones had to be present during heat inactivation to promote efficient renaturation upon temperature down-shift, suggesting that their critical function was in preventing the formation of insoluble protein aggregates rather than in their disaggregation. The data indicated that the mechanism of the reactivation reaction is very similar to that of the refolding of chemically denatured luciferase. DnaJ interacted first with luciferase and then formed the ternary complex with DnaK (135, 143). These results suggest that the capacity of DnaK, DnaJ, and GrpE to solubilize nonspecific protein aggregates may be limited to certain cases. More generally, the binding of DnaJ and DnaK to proteins as they begin to unfold under stress conditions may be more important in the prevention of aggregation.

Function of DnaK, DnaJ, and GrpE in De Novo Protein Folding

Evidence is accumulating that DnaK and DnaJ play an important role in de novo protein folding by interacting with nascent polypeptide chains. DnaK and DnaJ have been found to cofractionate with polyribosomes and to be released, together with the nascent polypeptides, by puromycin treatment (34). Cell-free translation reactions served to demonstrate that ribosome-bound chains as short as 55 amino acids bind DnaJ (58, 78). This interaction was shown to prevent subsequent folding unless DnaK and GrpE were also present in the reaction mixture.

A precise role for DnaK, DnaJ, and GrpE in the folding of newly synthesized polypeptides as part of a chaperone relay was proposed on the basis of in vitro reconstitution experiments (Fig. 4). These experiments were inspired by the finding that unfolded polypeptides entering the mitochondria from the cytosol interact with mitochondrial hsp70, a DnaK homolog, followed by hsp60, the homolog of E. coli GroEL, in order to fold (141). Langer et al. presented evidence that the folding of certain proteins may involve a two-step process. According to this model, the DnaK-DnaJ-GrpE system would transfer unfolded polypeptide to GroEL, which would then mediate its folding to the native state (89). When unfolded rhodanese, a mitochondrial enzyme, is diluted from denaturant into buffer, most of the protein aggregates. However, in the presence of DnaK, DnaJ, and ATP, a ternary chaperone complex forms in which rhodanese is stabilized against aggregation. In contrast to the observations made with firefly luciferase as the substrate (135, 143), addition of GrpE promoted only very inefficient folding of rhodanese, although the complex dissociated and reformed in the presence of ATP. However, when GroEL and its cofactor, GroES, were added (see below), very efficient refolding occurred, with a half-time of 30 min. The addition of GroEL and GroES to a reaction that lacked GrpE did not result in rhodanese refolding, suggesting that GrpE couples the DnaK- and DnaJ-mediated stabilization of an unfolded protein to the GroEL- and GroES-mediated refolding (89). Recently, homologs of DnaJ and GrpE have been discovered in mitochondria consistent with their obligatory role in protein folding outlined by the in vitro experiments with the E. coli proteins (141).

Fig. 4Model for the pathway of protein folding in the bacterial cytosol. As the polypeptide chain emerges from the ribosome, it is bound by DnaJ and DnaK (hsp70). The direct interaction between DnaK and DnaJ in the presence of ATP leads to the formation of a ternary complex between nascent chain, DnaK, and DnaJ in which DnaK is in the ADP state. This complex is resolved upon GrpE-dependent dissociation of ADP and the binding (not hydrolysis) of ATP (143). The protein may then fold to the native state by multiple rounds of interaction with the DnaK-DnaJ-GrpE system or may be transferred for final folding to GroEL-GroES (89). It is likely that a large fraction of cytosolic proteins have to interact with both chaperone systems to reach the native state.

Source:

Studies in vivo support the existence of a cellular folding pathway in which DnaK, DnaJ, and GrpE cooperate with GroEL and GroES in the bacterial cytosol. Overproduction of either DnaK-DnaJ or GroEL-GroES in rpoH mutants, which lack the heat shock factor σ ³² and thus are almost completely devoid of heat shock proteins, prevented the otherwise inevitable aggregation of newly synthesized proteins (43). An rpoH revertant strain, selected at 20°C, was able to constitutively express GroEL and GroES, but a significant degree of newly synthesized protein aggregation persisted in this strain. A second revertant, selected at 30°C, was able to express both GroEL-GroES and DnaK-DnaJ. In this strain, aggregation was almost completely suppressed, suggesting that both of these chaperone systems collaborate in folding.

GroEL- and GroES-Mediated Protein Folding

The so-called chaperonin GroEL and its regulator GroES (sometimes called cochaperonin) (56) are required for the de novo folding of a large percentage of cytosolic proteins (65) and participate in related processes such as protein translocation and degradation (11, 84, 139). Homologs of GroEL and GroES are found in mitochondria and chloroplasts, reflecting the origin of these organelles from endosymbiotic bacterial ancestors. In contrast, the chaperonin of the eukaryotic cytosol, known as the T-complex polypeptide (TCP)-ring complex, although similar in function and general architecture, bears little sequence relationship to GroEL (77, 93). Both GroEL and GroES, while heat inducible, are constitutively expressed and are essential for growth under all conditions (27). GroEL was first identified as a host protein required for the assembly of bacteriophage λ heads and T5 tails (38). However, the first indication that the chaperonins were involved in protein assembly came from studies with the enzyme ribulose bisphosphate carboxylase-oxygenase (Rubisco) in plant chloroplasts. Hemmingsen and Ellis (55) demonstrated that the large subunits of Rubisco formed a complex with a "binding protein" prior to their assembly into the holoenzyme and that the release of the large subunit was ATP dependent (for a review, see reference 25). The "binding protein" turned out to be the chloroplast homolog of GroEL (56). A second study by Cheng et al. described a mutation in the yeast mitochondrial homolog of GroEL, hsp60, that resulted in defective assembly and aggregation of several proteins following their import from the cytosol into the mitochondria (17, 18). Ostermann et al. established that the primary role of the mitochondrial hsp60 is not to assemble multimeric proteins; rather, its function is to mediate polypeptide chain folding in an ATP-dependent reaction, a process that had previously been assumed to occur spontaneously (114).

As in the case of mitochondrial hsp60, the general role of GroEL in folding was demonstrated in an E. coli strain conditionally defective in GroEL (65). Loss of GroEL function under controlled conditions results in the aggregation or rapid degradation of many newly synthesized polypeptides. The interaction of GroEL with newly translated proteins has been demonstrated in vivo (34) and in cell-free translation reactions (11). From these studies, it appears that the chaperonin interacts primarily with complete polypeptide chains after their release from the ribosome. The ability of GroEL to interact with many different proteins in their nonnative conformation has been established convincingly on the basis of a large number of individual case studies in vitro (26, 42, 44, 46, 103, 107, 119, 148). At least 50% of the soluble proteins of E. coli formed complexes with GroEL when diluted from denaturant into GroEL-containing solution (151). Notably, proteins in their native states did not interact with GroEL. In light of the broad substrate specificity, the capacity of the chaperonins to distinguish precisely between nonnative and native structures is certainly one of their most fascinating functional features. Similar results were obtained for mitochondrial hsp60 in organello. When yeast mitochondria were exposed to heat shock temperatures, a wide range of proteins were found in association with the chaperonin, probably reflecting a role of the chaperonin system in stabilizing preexistent proteins under stress conditions (102). GroEL overexpression was found to suppress the phenotypes of temperature-sensitive mutations in genes encoding several diverse proteins of Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) (150).

The function of GroEL and GroES has been reconstituted in vitro with chemically denatured substrate proteins, demonstrating initially the assembly of prokaryotic Rubisco (41) and then the folding of monomeric proteins such as dihydrofolate reductase and rhodanese (103, 107). Studies with ornithine transcarbamylase, a homotrimeric enzyme of mitochondria, as a substrate have demonstrated that GroEL and GroES participate actively in the chain folding of the subunits, which can subsequently assemble spontaneously into the active trimer (161).

Structure of GroEL

Very few studies of protein structure have revealed as much about their mechanism of action as they have for the chaperonins. GroEL is a tetradecameric complex of identical 57-kDa subunits. Early electron microscopic analysis revealed its double toroidal structure, composed of two stacked heptameric rings with approximate dimensions of 14 nm in diameter and 16 nm in height (59, 62). This peculiar symmetry has been confirmed many times (70, 90, 162), but, most recently, Saibil et al. performed a high-resolution electron-microscopic analysis of the hsp60 from Rhodobacter sphaeroides (131). This analysis enabled the clear visualization of the two seven-membered (7-mer) rings and supported the long-held view that the hsp60 monomer consisted of two major domains. The domains of each monomer are linked on the exterior of the cylinder, giving hsp60 a cagelike structure (131). The central cavity of the cylinder, intuitively the site of polypeptide binding, is 4 to 6 nm in diameter and can expand considerably upon binding of the GroES cofactor (see below). The enormous task of solving the structure of the GroEL tetradecamer by X-ray crystallography has recently been accomplished by Braig et al. (13) (Fig. 5). Each monomeric unit is composed of an apical, an intermediate, and an equatorial domain. The equatorial domain is the largest of the three and provides the majority of the subunit-subunit interactions and all of the interactions between the rings across the equatorial plane of the cylinder. The N- and C-terminal residues protrude into the cavity from the equatorial domain; however, they are disordered and are not crystallographically resolved. Mutagenesis has demonstrated that this domain contains the ATP-binding site (28). The intermediate domain is the smallest of the three and provides the connection between the other two. The connection between the intermediate domain and the other two domains is mediated through two short antiparallel segments which could easily serve as hinges in allosteric conformational changes. The apical domain forms the opening of the central channel. The regions of this domain that face the channel are relatively disordered, probably reflecting conformational flexibility. Mutational analysis implicates these regions in polypeptide binding (28).

Fig. 5Schematic models of the GroEL structure. (A) Principal dimensions of the GroEL double toroid determined from the three-dimensional crystal structure. The top panel shows a view down the central cavity, while the bottom panel represents a vertical cut through the cylinder. (B) Domain structure of GroEL. A, I, and E correspond to the apical, intermediate, and equatorial domains, respectively; C corresponds to the central cavity; W corresponds to an external opening of a side window.

Source: Reprinted from reference 13 with permission.

GroES is a single heptameric ring of identical 10-kDa subunits (14, 53, 101). The stoichiometry of GroES 7-mer binding to GroEL 14-mer has been demonstrated to be 1:1 by both biochemical experimentation and electron microscopy (90). (Under certain conditions, for example, in the presence of ATP, at high levels of magnesium and at slightly alkaline pH, both ends of the GroEL cylinder may bind GroES [3, 100, 134]. However, the significance of these "football-shaped" particles remains to be determined, because only the asymmetrical complex is populated under most conditions in which GroEL is fully functional.) The contact with GroEL is made through a flexible loop region in each GroES subunit (88). Electron microscopic images of the complexes show one GroES 7-mer sitting asymmetrically on only one side of the GroEL double toroid (90, 130, 131). Comparison of side-on views of GroEL alone with those of the GroEL-GroES complex shows a degree of distortion in the GroES distal ring of GroEL, suggesting that GroES binding induces a conformational change in this ring that prevents the association of a second GroES. End-on views of GroEL with or without bound nucleotide indicate that a considerable conformational change occurs in the ring structure upon nucleotide binding (90). ATP binding induces a change in the orientation of the outer domain with respect to the inner domain of each GroEL subunit (131). This is even more dramatic upon asymmetrical binding of the GroES cofactor (16). The apical domains of the GroEL subunits that interact with GroES open outward, forming an enclosed dome-like space with a maximum height of 65 Å (6.5 nm) and a maximum width of 80 Å (8.0 nm) (Fig. 6). It has been speculated that this enclosed space could transiently accommodate a folding polypeptide chain.

Fig. 6Structure of the GroEL-GroES complex. Shown is a three-dimensional reconstruction of the structure of the GroEL-GroES-ATP complex as determined by cryoelectron microscopy. The structure clearly shows the extensive opening out of the GroEL apical domains on the side to which GroES is bound.

Source: Reprinted from reference 16 with permission.

Polypeptide Binding by GroEL

GroEL binds one or two molecules of unfolded polypeptide (10, 103, 106). The central cavity of the cylinder was a good candidate for the substrate-binding site, since it could potentially accommodate even relatively large globular proteins. Evidence for this hypothesis was again derived from electron microscopic images of GroEL to which the substrate protein rhodanese had been bound. End-on views of these negatively stained GroEL molecules revealed an additional mass in the central cavity that was not seen in any of the images of GroEL in the absence of substrate protein (90). Unfolded gold-labeled dihydrofolate reductase was localized to the central cavity, possibly in association with the outer GroEL domains (12). Cryoelectron microscopy of a GroEL-GroES complex to which ATP and unfolded malate dehydrogenase had been bound demonstrated that the substrate was localized inside the GroES-distal ring of GroEL (16). Mutagenesis experiments have confirmed that the substrate makes contact with these regions of the channel surface (28), which contain a number of exposed hydrophobic amino acid residues that probably formed a flexible hydrophobic binding surface.

What is the structural element on a putative substrate that GroEL recognizes, and what is the conformation of the polypeptide once it is bound to GroEL? Answers to these questions began to emerge from the study by Martin et al. in which it was proposed that GroEL bound the substrate protein in a molten-globule-like conformation (103). This was based on the analysis of the intrinsic tryptophan fluorescence of both dihydrofolate reductase and rhodanese while in a complex with GroEL and on the fact that these proteins bound the hydrophobic fluorescent probe 1-anilino-naphthalene-8-sulfonate (ANS) which is diagnostic for these partially folded states (103). The fluctuating hydrophobic surfaces exposed by the compact molten-globule state are likely to interact with the hydrophobic binding surfaces identified in the GroEL channel.

Recent studies, using advanced biophysical techniques, have indeed shown that polypeptide binding by GroEL is driven by hydrophobic forces and have confirmed the nature of the flexible, yet compact protein conformation that is recognized by the chaperonin (54, 129, 160). A study of the conformational specificity of GroEL for compact intermediates of α-lactalbumin indicated a preference for molten-globule-like structures (54). Hydrogen-deuterium exchange as measured by NMR has been applied to investigate the conformation of cyclophilin as a complex with GroEL. Cyclophilin has 39 backbone amide protons that are nonexchangeable in the native protein. However, if the protein is bound to GroEL, all of these 39 amide protons can be exchanged for deuterium, suggesting that the cyclophilin intermediate bound to GroEL is in a conformation in which all of the secondary and tertiary structural elements are destabilized, a description that fits well within the definition of the molten globule. Again by using deuterium exchange techniques, it was shown for α-lactalbumin that the secondary structure in the bound polypeptide has very similar stability to that of the molten-globule form free in solution (129).

GroEL can interact with the hydrophobic face of an amphiphilic α-helix, thus potentially stabilizing the α-helical structure in the bound protein (86). This is in contrast to the chaperones of the hsp70 family, which bind peptide sequences in an extended conformation (87). Moreover, GroEL does not form stable complexes with small peptides. Stable binding requires larger polypeptide segments of perhaps 50 to 100 residues, which are capable of forming hydrophobic surfaces by folding into a compact conformation. The interaction is apparently independent of the presence of α-helical elements, because all-β-structured proteins can also bind GroEL (132). It appears, therefore, that what the chaperonin recognizes are the flexible hydrophobic surfaces typically present in partially folded proteins independent of specific secondary structure elements.

GroEL/GroES Reaction Cycle

The precise mechanism of GroEL function is still under active investigation, but certain basic principles have emerged (Fig. 7). Most of the mechanistic studies have been carried out in vitro with pure proteins. GroEL has a weak ATPase activity (59, 71, 72), and ATP binding and hydrolysis regulate the interaction of unfolded polypeptide with the chaperonin. GroES functions as a regulator, coupling ATP hydrolysis by GroEL to productive folding (14, 45, 103, 152). Interestingly, while a variety of denatured proteins may bind GroEL, only those proteins which are unable to reach the native state spontaneously in vitro depend critically on both GroES and ATP hydrolysis for their folding by GroEL (103, 133). This has been shown with the mitochondrial protein rhodanese, which has a strong tendency to aggregate when refolding is attempted in the absence of chaperonin. These findings led to the proposal that critical folding steps occur in association with the central cavity of GroEL in a reaction that requires the cooperation of GroES (103, 104). Thus, the polypeptide would be discharged into the solution only after it has buried a significant amount of hydrophobic surface, thereby reducing its tendency to aggregate. Among other findings, this model is supported by the observation that ATP-dependent release of rhodanese from GroEL in the absence of GroES results in aggregation and that GroEL is apparently able to interact with different conformational states of the protein (103). Furthermore, certain small proteins such as barnase are apparently capable of folding on the chaperonin surface (46). An alternative view has been put forward by Horwich, Lorimer, and coworkers on the basis of recent experiments demonstrating that a single round of ATP hydrolysis by GroEL releases bound protein in a not yet native conformation that subsequently may rebind to another chaperonin molecule (146, 157). These authors proposed that the chaperonin essentially releases the substrate protein in an unfolded state for spontaneous folding in bulk solution. Protein molecules that do not reach the native state upon a single release cycle and rather misfold will rebind to the chaperonin, where they are unfolded to the initially bound state in preparation for another trial of spontaneous folding. Thus, here the protein is released from the chaperonin in a more unfolded state than that bound originally. Both models concur in the finding that multiple rounds of ATP-dependent release and rebinding are necessary for complete folding (103, 157). However, the model by Martin et al. favors the view that rebinding occurs preferentially to the same chaperonin molecule and that signficant folding occurs in the protected environment of the chaperonin before the protein reaches the bulk solution. This model also includes the possibility that a structural rearrangement of unproductive folding intermediates occur upon rebinding to GroEL (57, 103). Formal proof that a polypeptide can fold all the way to the native state within the chaperonin cavity has yet to be presented.

Fig. 7Model for the ATP-dependent interaction between GroEL, GroES, and folding polypeptide. 1. Binding of polypeptide substrate facilitates the dissociation of GroES by accelerating ATP hydrolysis in the substrate-bound ring of GroEL and by lowering the affinity of ADP in the opposite ring. 2. Polypeptide is bound stably in the nucleotide-free state of GroEL. 3. ATP and GroES rebind. GroES may associate either with the free GroEL ring (step 3a) or with the polypeptide-containing ring (step 3b). 4a. ATP hydrolysis leads to polypeptide release and incompletely folded polypeptide rebinds in (step 1). In step 3b, polypeptide is transiently enclosed in the central cavity and is free to fold. ATP hydrolysis in the GroES-bound toroid then generates the tight ADP state (not shown). ADP dissociates upon ATP hydrolysis in the opposite ring, causing dissociation of GroES and allowing polypeptide release. ADP, the high-affinity ADP state of the seven subunits of GroEL which are bound to GroES; ADP, the lower-affinity ADP state in the opposite toroid which may hydrolyze ATP; ATP, the subunits in a GroEL toroid in the ATP-bound state; I, polypeptide substrate as a compact folding intermediate; I*, folding intermediate partway advanced toward the native state; N, native protein.

Source:

In contrast to the actual mechanism of folding, the cooperation between GroEL and GroES, as well as the nucleotide exchange reactions that occur during a folding reaction, is well understood. The GroEL ATPase activity is dependent on K⁺ ions (152) and is stimulated upon binding of unfolded polypeptide substrates (73, 104). In the presence of GroES and sufficiently high levels of potassium, the ATPase activity of GroEL is inhibited by 50% but its cooperativity is increased so that one GroEL toroid hydrolyzes seven ATP molecules at a time (14, 45, 73, 90, 103, 145, 152). Complex formation between GroEL and GroES is nucleotide dependent. GroES is thought to initially bind to a GroEL toroid in the ATP state. This is followed by ATP hydrolysis in that toroid and results in a very stable GroEL-GroES complex with seven ADP molecules tightly bound to the GroES-interacting subunits of GroEL (104). ATP hydrolysis in the GroES-bound toroid is then inhibited, explaining the half-of-the-sites inhibition reflected in the 50% reduction of the GroEL ATPase. Interestingly, the interaction between GroEL and GroES is highly dynamic (104). The complex dissociates when ATP is hydrolyzed in the GroES-distal ring of GroEL, allowing GroES to cycle between bound and free states.

How is this basic interaction between GroEL and GroES modulated upon binding of substrate polypeptide by GroEL? Polypeptide binding into the free ring cavity of the chaperonin has been shown to accelerate the dissociation of GroES from the opposite GroEL toroid (104) (Fig. 7). Rebinding of GroES followed by ATP hydrolysis is then required to trigger a cycle of protein release for folding. In principle, GroES could exert its effect in facilitating polypeptide release from the GroEL toroid distal to the substrate-bound toroid by a long-range allosteric effect. Alternatively, rebinding of GroES to the GroEL-polypeptide complex may occur to the same toroid that contains the substrate protein. This possibility seems attractive in light of the increase in volume of the central cavity observed upon GroES binding (see above) (16). Substrate protein would then probably be displaced into the cavity, where folding could initiate while the ring opening is closed by GroES, preventing premature exit of the polypeptide. GroES would dissociate upon another round of ATP hydrolysis in the opposite toroid, potentially releasing the substrate into the solution. At present, it seems possible that both mechanisms of GroES-mediated polypeptide release from GroEL coexist (50).

CHAPERONE-DEPENDENT PROTEIN TRANSLOCATION

A large number of the proteins synthesized in the E. coli cytosol have to be translocated into and across the plasma membrane (159). This process is mediated by specific targeting sequences at the amino terminus of the proteins and by a specific protein machinery comprising a set of integral and peripheral membrane proteins as well as chaprerone components in the cytosol (see chapter 64). It had been demonstrated in the late 1970s that several export proteins are elongated on membrane-associated polyribosomes (125, 138); however, it was found that translocation initiates relatively late in translation, an important distinction from protein secretion into the endoplasmic reticulum in mammalian cells. For maltose-binding protein (MBP) and ribose-binding protein (RBP), two soluble enzymes of the peri-plasmic space, it was demonstrated that translocation initiated only after 80% of the protein had been synthesized in the case of MBP or after the protein was fully elongated in the case of RBP. The general view is that co- and posttranslational modes of translocation coexist in E. coli.

Requirement of an Unfolded State for Translocation

From the analysis of different membrane systems in bacteria and eukaryotic cells, it is clear that translocation generally requires the polypeptide chain to be in an unfolded state (24, 159). If a protein can be fully translated prior to translocation, it might in principle fold in the cytoplasm before making contact with the translocation machinery in the membrane. Classical experiments by Randall and Hardy established that when such folding takes place, export to the periplasm is prevented (126). In these experiments, the kinetics of folding of export-competent, wild-type pre-MBP was compared with that of a mutant form which could not be exported as a result of a defective leader sequence. When the experiment was performed under conditions where export was blocked by an uncoupler of the membrane potential, wild-type pre-MBP remained in a protease-sensitive, loosely folded form whereas the mutant precursor rapidly folded into a protease-resistant, stable conformation. Interestingly, the kinetics of folding of the mutant pre-MBP closely paralleled the kinetics of the loss of translocation competence, measured in the absence of uncoupler. This observation allowed not only the conclusion that the folding of precursor proteins precludes them from undergoing translocation but also that precursors must somehow be maintained in an unfolded conformation prior to export (126).

Function of the Chaperone SecB

How are precursor proteins maintained in a translocation-competent form for export? E. coli contains a chaperone protein, SecB, devoted to this function (127). The existence of such a factor was predicted from experiments studying an export-defective mutant of MBP whose expression severely blocked the translocation of wild-type pre-MBP and another exported protein, pro-OmpA (6, 7). These experiments suggested that the interference was due to the mutant MBP titrating out the factor required for the maintenance of the translocation-competent conformation of precursor proteins. Similarly, the study by Randall and Hardy described above also suggested the existence of such a factor (126).

The disruption of a new gene, secB, was demonstrated to inhibit the translocation of a number of outer membrane proteins (82) at an early stage of the translocation process (83). Subsequently, it was demonstrated that SecB was indeed the factor that was depleted by the overexpression of mutant MBP (20). All of the proteins shown to be SecB dependent for export were also sensitive to the overexpression of the mutant MBP. Significantly, translocation of wild-type pre-MBP was fully restored when SecB was overexpressed, clearly demonstrating that SecB was responsible for maintaining the translocation competence of a subset of precursor proteins (20). If the interaction with SecB retarded pre-MBP folding, the complexed pre-MBP should be susceptible to proteinase digestion. This was indeed shown to be the case. By using an E. coli translation lysate derived from a SecB⁻ strain, pre-MBP was found to fold rapidly into a proteinase resistant conformation. However, if a SecB⁺ strain was used, a large proportion of the pre-MBP was susceptible to proteinase K digestion. This effect was even more pronounced when the lysate was prepared from the SecB-overproducing strain (20). The direct physical interaction between SecB and precursor proteins was finally demonstrated by the isolation of a complex between SecB and pro-OmpA by gel filtration chromatography (91). On the basis of these observations, SecB can be defined as a chaperone that binds to a number of newly synthesized secretory proteins, maintaining them in the conformationally unfolded state required for translocation across the periplasmic membrane.

SecB is a tetramer composed of identical 16-kDa subunits (156). The protein has been purified, allowing the reconstitution of its function in vitro. While SecB is able to interact with the signal sequences of secretory proteins (156), it has been clearly demonstrated that many binding sites reside within the mature regions of precursor proteins (20, 35, 99). Four molecules of peptide ligand can bind per SecB tetramer (124). Peptide-binding assays were used to establish the structural features recognized by SecB in its substrates. Interestingly, the peptides that bound best were positively charged and flexible in conformation. Peptide binding rendered SecB resistant toward proteolytic cleavage of its C-terminal ∼50 amino acid residues (124). Since the same protease resistance was observed when SecB was exposed to high ionic strength (124), this suggested that SecB undergoes a significant conformational change upon peptide binding.

The recognition of its substrate via a charge-charge interaction distinguishes SecB from other chaperones such as GroEL or hsp70. However, SecB also possesses hydrophobic binding sites, which probably interact with the hydrophobic regions exposed by unfolded precursor proteins. The existence of such sites was demonstrated by using the hydrophobic fluorescent dye ANS, which changes its fluorescence properties when associating with a hydrophobic protein surface (124; also see above). SecB showed significant ANS binding only upon complex formation with positively charged peptide. In light of these results, the initial binding of substrate protein to SecB is thought to be mediated via an electrostatic interaction. Multiple occupation of such sites might subsequently induce a conformational change, causing the exposure of the hydrophobic peptide-binding region of SecB, which would then promote a tight interaction with additional, hydrophobic parts of the substrate protein. Support for this hypothesis was obtained by characterizing the sequences within pre-MBP to which SecB binds. Three distinct, nonoverlapping regions were identified, two of which were predominantly positively charged. However, all three peptides were in proximity to stretches of hydrophobic sequences (147). This dual recognition would allow SecB to faithfully discriminate between unfolded and folded proteins. Only with unfolded polypeptides, such as nascent precursor proteins, would the initial interaction with charged residues probably be followed by an interaction with exposed hydrophobic regions (124).

While it seems clear how SecB differentiates between folded and unfolded, it is less clear how a distinction is made between a nascent protein destined for export and one which is to reside in the cytoplasm. Clearly, the presence of a leader sequence is characteristic of secretory proteins, and these sequences combine the two structural features recognized by SecB, positive charges followed by a hydrophobic sequence (153). However, SecB seems to interact primarily with sequences from the mature region of precursor proteins. Significantly, it was observed that the presence of a leader sequence lowers the rate of folding of precursor proteins relative to that of their mature counterparts (85, 117). The leader sequence could indirectly mediate the interaction of a secretory protein with SecB by kinetically retarding its folding. Experimental evidence has been presented that slowing the spontaneous folding of MBP enhances its capability to interact with SecB in vitro, even in the absence of the leader sequence (49). Thus, the rate of folding of a nascent protein is believed to strongly influence its ability to bind to SecB in vivo, and the fate of the protein is thought to be determined by a kinetic partitioning between the folding pathway of a nascent protein and its association with SecB (49).

How is the complex between SecB and the precursor eventually resolved to allow translocation? Unlike hsp70 or GroEL, SecB does not bind and hydrolyze ATP. Interestingly, it has been shown that SecB interacts directly with a second protein of the secretory machinery, SecA. SecA is a peripheral membrane component, and its ATPase activity is required to drive protein translocation (see chapter 64). The interaction of SecB with SecA may guarantee the ordered transfer of the precursor proteins to the translocation machinery in the plasma membrane (51, 142).

PERIPLASMIC CHAPERONES

Following their membrane translocation, the folding and assembly of proteins are generally believed to depend on chaperone proteins and other factors in the target compartment. This is well understood for eukaryotic organelles, such as mitochondria, chloroplasts, and the endoplasmic reticulum. In contrast, comparatively little is known about chaperone proteins in the bacterial periplasm that may facilitate the folding of exported proteins. The best-studied periplasmic chaperone, PapD, participates in the asssembly of pili. It is these structures at the surface of gram-negative bacteria that contain at their tips adhesive molecules that mediate the process of microbial colonization and infection. Most uropathogenic strains of E. coli isolated from humans express P pili which bind to specific sugar residues present in the glycolipids on the surface of the cells lining the upper urinary tract (67).

All of the proteins involved in the synthesis and assembly of P pili are encoded on two pap gene clusters, which code for 11 independent proteins (66, 67). P pili contain two distinct structures, a stalk (pilus shaft) and a tip fibrillum (79). The stalk is made up of repeating units of PapA. The tip fibrillum is composed of PapK, which makes the connection with the stalk, PapE, which is the major component of the tip fibrillum, and PapF, which is the adapter linking PapG to the fibrillum. PapG contains the sugar-binding site (67). Two other proteins that are also encoded by the pap operon are PapC and the chaperone PapD. Both are localized in the periplasmic space. PapD is a soluble protein, while PapC is associated with the outer membrane. The following observations characterized PapD as a "private" molecular chaperone of the pilus system. A complex of PapD and PapG was isolated in an attempt to purify PapG by affinity chromatography on galabiose-Sepharose (69). Complex formation with PapD was found to render PapG resistant to proteolysis and to enhance the processing of its signal peptide during export from the cytosol. A genetic lesion in papD resulted in the proteolytic degradation of PapA, PapE, PapF, and, to a lesser extent, PapG (97). Significantly, the interactions between PapD and pilus components were limited to the assembly process, because PapD was not detectable in the final macromolecular pilus structure.

Structure and Function of PapD

PapD was the first chaperone whose three-dimensional structure was solved by X-ray crystallography (63). The protein is composed of two domains which are oriented in such a way that the molecule has the shape of a boomerang. Each of the domains consists of an antiparallel β-barrel formed by two β-sheets packing against each other. Of particular interest is the topology of the domains, which is very similar to that of immunoglobulin constant domains. Further analysis of the structure suggested that a region in the cleft between the two domains might form the binding site for pilus components. The central part of this region contains solvent-exposed hydrophobic residues flanked by three basic residues on the N-terminal domain and three acidic residues on the C-terminal domain (63). The PapD sequence was aligned with a number of other PapD homologs, and the derived consensus sequence was superimposed onto the PapD structure (64). The invariant residues are located predominantly on the β-strands in the cavity between the two domains. Most of these residues are involved in maintaining the overall fold of the domains or their relative orientation. However, a number of residues are invariant and have been shown by site-directed mutagenesis to be critical for substrate binding. Alignment of PapD substrate proteins has demonstrated a high degree of conservation in their carboxy termini (113). Binding of PapG to PapD seems to be mediated by the 14 C-terminal residues of PapG, because removal of this segment prevented the formation of the PapG-PapD complex in vivo (69). Interestingly, the PapD-interacting sequences of PapE, PapF, and PapK also appear to be localized at the C termini of these proteins (81).

Detailed insight into the structural basis for substrate binding to PapD arose from the crystal structure of a PapD-PapG peptide complex (81). The peptide is bound in an extended conformation along the cleft β-strand of domain 1 of PapD, while its C-terminal residue (proline) is firmly anchored within the cleft between the two domains (Fig. 8). The major stabilizing interaction is provided by hydrogen bonding, but hydrophobic interactions also provide significant stabilization energy (81). Of the 19 residues from the PapG peptide bound to PapD, 5 make direct contacts with PapD. Of these, Phe-2, Leu-4, Met-6, and Met-8 are part of the conserved pattern of alternating hydrophobic and hydrophilic residues found in the C termini of many of the pilus subunits. Similarly, Leu-103, Ile-105, and Leu-107 in PapD form a pattern of hydrophobic residues that is conserved among the members of this family of periplasmic chaperones.

Fig. 8Structure of the PapD-PapG peptide complex. The PapG peptide is bound in an extended conformation along the G1 ?-strand in the cleft of PapD. The principal binding interactions are shown and discussed in detail in the text.

Source: This figure was a gift from S. Hultgren (81).

An important difference between PapD and the cytoplasmic chaperones seems to lie in the conformation of their respective substrates. We have already seen that the cytoplasmic chaperones bind to extended polypeptide chains or to compact folding intermediates exposing hydrophobic surfaces. PapD, on the other hand, appears to bind its substrates as native-like, assembly-competent intermediates, as judged by the ability of PapG to recognize its receptor while it is bound to PapD (80). However, as with some of the cytoplasmic chaperones, the PapD substrate is in rapid exchange between bound and free forms. This is based on the finding that ¹²⁵I-PapD added to a preformed PapD-PapG complex becomes incorporated into the complex. When the complex was dissociated in denaturant and then rapidly diluted into buffer, all of the PapG aggregated. The presence of PapD in the dilution buffer prevented aggregation by sponsoring the reformation of the complex. Therefore, it seems likely that the interaction with PapD masks a site on PapG which renders the protein prone to aggregation but which, when released at the appropriate site for assembly, allows the subunit to assemble into the growing pilus (80). Thus, while the majority of the PapD-associated protein may be in a folded conformation, the structural element recognized by PapD is flexible and nonnative like. The situation may be comparable to the interaction of DnaK and DnaJ with folded protein components in λ DNA replication in which an interaction that generally occurs with unfolded polypeptides is utilized for a specific regulatory purpose. In the case of PapD, however, the chaperone recognition element seems to be limited to the pilus components.

PapC, a "Molecular Usher"

PapD cooperates in pilus asembly with PapC, an outer membrane protein that has been called a molecular usher. A lesion in PapC results in the accumulation of complexes of PapD with unassembled pilus subunits in the periplasm (112). This suggested that PapC is the site in the outer membrane to which PapD-substrate complexes are targeted for assembly. Evidence supporting this view came from studies investigating the interaction of PapD with PapC in vitro. PapD alone was not able to form a complex with PapC; however, complexes of PapD with either PapG, PapE, or PapF were able to bind to PapC (23, 140). By using a pulse-chase protocol, direct evidence was obtained that PapD-PapA complexes are the true assembly intermediates that are targeted to PapC for incorporation into the growing pilus shaft. Thus, PapD acts as a chaperone whose specific functional role is to assist in the assembly of P pili by binding to their structural subunits as they are translocated into the periplasm. Stabilized by PapD, the subunits are then handed over to membrane-bound PapC in a native-like, assembly-competent conformation, where they are preferentially released and incorporated into the growing pilus.

CONCLUSIONS

On the basis of their exquisite ability to distinguish between native and nonnative protein structures, molecular chaperones participate in a number of cellular processes that depend critically on the conformation of the proteins involved. Proteins destined to fold in the cytosol are chaperoned so that their folding is efficient and the formation of misfolded protein aggregates is minimized. For assembly into oligomers, protein subunits must be folded into native-like conformations that are assembly competent. Chaperones are involved in stabilizing such intermediates. Precursor proteins to be translocated across the periplasmic membrane are maintained by chaperones in an unfolded conformation. The capacity of chaperones to stabilize nonnative states may also be utilized in the pathways of protein degradation. This is suggested by a number of recent studies (75, 139, 154) discussed in detail in chapter 62. Finally, in the periplasm, a subset of proteins requires the assistance of chaperones in order to be delivered to specific assembly sites on the outer membrane. The underlying principle in most of these reactions is the interaction of chaperones either with extended peptide segments enriched in hydrophobic amino acid residues or with hydrophobic surfaces exposed by partially folded substrate proteins. Once the substrates are bound, specific chaperone classes have different mechanisms by which they productively release them. These general principles may be adapted to specific regulatory purposes where the interaction with chaperones alters the activity state of key components.

The further elucidation of the reaction mechanisms of chaperones will shed light on the similarities and differences between the in vivo and in vitro mechanisms of protein folding. Such information will be derived from detailed structural analysis, allowing the design of rational mutagenesis strategies. For GroEL, PapD, and, to some extent, hsp70, such studies have begun to emerge. More detailed investigations will directly provide information describing precisely the interactions of the substrate proteins with the chaperones and how they are guided into their final conformation. Of similar importance is the cooperation between different chaperones and also between chaperones and other protein-folding catalysts (disulfide isomerases and prolyl isomerases) in cellular pathways of protein folding (for reviews, see references 40 and 110). Clearly, the structural and functional analysis of the cellular machinery that mediates protein folding will remain an active area of research for many years to come.