Anfinsen comes out of the cage during assembly of the bacterial pilus

The classic experiments of Anfinsen and his associates established that the primary sequence of a polypeptide contains all of the information needed for a polypeptide to fold into its native fully active tertiary structure (1). Although Anfinsen's conclusions still hold, for the most part, it is now becoming clear that certain proteins, such as the subunit proteins building up adhesive surface organelles on Escherichia coli, need additional steric information from other proteins in order to fold properly (2). In their paper published in this issue of PNAS, Barnhart et al. (2) show that at least some proteins may require information for folding to be transiently provided by distinct molecular chaperones.

There are many chaperones involved in protein folding; however, few are known to act by directly donating steric information to their substrate proteins. For example, the GroEL-GroES chaperone system couples ATP hydrolysis to the iterative binding and release of unfolded polypeptides. The binding stabilizes the substrate in an unfolded state until it is released and allowed to fold. The GroEL (3) chaperone is a multimeric ATP-driven macromolecular complex that has been proposed to facilitate protein folding via the Anfinsen cage model (4, 5). This model is based on the view that protein folding in vivo is limited by intermolecular reactions that produce aggregation. It proposes that the GroEL cavity provides a sequestered microenvironment in which folding to the native state can proceed while the substrate is protected from aggregation. Proteins that require the GroEL chaperonin for in vivo folding do not receive steric information from this class of chaperones. Rather, these chaperones act to prevent or overcome the misfolding of their substrate polypeptides. The DnaK (Hsp70) family represents another major class of cytoplasmic chaperones (6, 7). In conjunction with the co-factors DnaJ and GrpE, DnaK couples ATP hydrolysis with substrate binding and release. Similar to the GroEL system, the DnaK family members do not contribute steric information to their substrate proteins. DnaK is thought to bind to unfolded proteins via exposed hydrophobic residues, preventing their aggregation and misfolding, until they are released into the cytoplasm, where folding is completed (6, 7). The substrates for DnaK can be either newly synthesized proteins or proteins misfolded because of stress conditions.

In contrast to this cytoplasmic general folding machinery described above, recent work suggests that the small periplasmic PapD-like chaperones, which participate in the assembly of adhesive surface structures in many Gram-negative pathogens, facilitate folding by directly providing steric information to their substrates. PapD was the first chaperone for which the crystal structure was solved (8). It contains two immunoglobulin-like (Ig) domains oriented in a boomerang shape. PapD specifically interacts with pilus subunits before their incorporation into P pilus polymers in uropathogenic strains of E. coli (9, 10). The protein was characterized as a chaperone because it is required for pilus formation but does not appear in the pilus itself. PapD is one chaperone in a family of over 30 members that are involved in the assembly of adhesive structures on the bacterial cell surface. Two recently published crystal structures of PapD-like chaperones in complex with pilus subunits, or pilins, reveal the basis for their interaction with subunits (11, 12). Pilus subunits lack the C-terminal seventh β-strands that would otherwise complete their Ig folds. The missing strand results in a groove along the surface of the pilin that exposes its hydrophobic core. In the structures, the chaperone donates a β-strand (strand G₁) that occupies the groove and completes the fold of the subunit (donor strand complementation) (Fig. 1A). In the absence of the chaperone, subunits are unstable and are degraded by proteases such as DegP, which recognizes denatured, misfolded, and/or aggregated proteins. Donor strand complementation thus allows the subunit to correctly fold by providing steric information in the form of the missing strand. The subunit groove occupied by the G₁ strand also participates in subunit–subunit interactions in the pilus. Thus, donor strand complemention couples the folding of the subunit with the capping of its interactive groove, ensuring that the groove is never free to interact nonproductively in the periplasm. During pilus assembly, which occurs at the outer membrane usher, the N-terminal extension of an incoming subunit displaces the chaperone G₁ strand and occupies the groove of the most recently incorporated subunit (donor strand exchange) (Fig. 1B). The mature pilus thus consists of an array of subunits, each of which contributes a strand to complete the fold of its neighbor. Pili are heteropolymeric structures containing a number of different subunit proteins assembled in a specific order. The affinity of the incoming donor strand for the exposed hydrophobic cleft in the preceding subunit may in part determine the order of subunit incorporation. Pilus assembly by PapD-like chaperones thus represents a distinct variation of the Anfinsen principle, because subunits require information from another polypeptide entity—the chaperone and then the neighboring subunit—to attain and maintain, respectively, their proper fold. In these regards, PapD-like chaperones function similarly to intramolecular chaperones (IMCs), covalently attached peptides that have been shown to facilitate protein folding by providing steric information and that are subsequently cleaved from their substrates (13, 14). However, unlike IMCs, PapD-like chaperones are separate, fully folded proteins that provide steric information for pilin folding while simultaneously capping their interactive surfaces.

Figure 1.

Pilin domain topology diagrams. Dashes indicate additional polypeptide not shown. (A) In donor strand complementation, the chaperone contributes its G₁ strand (red) to complete the immunoglobulin-like fold of the subunit (white). The completed fold is noncanonical because the G₁ strand runs parallel to the subunit C-terminal F strand. The N-terminal extension is shown as a blue strand. (B) After donor strand exchange, the N-terminal extension of one subunit completes the Ig fold of its neighbor in a canonical manner, as the N-terminal extension runs anti-parallel to the F strand. (C) Donor-strand-complemented FimH (dscFimH) was constructed by fusing the N-terminal extension of FimG (blue), which is predicted to complete the fold of FimH in the pilus, to the C terminus of FimH with a 4-amino-acid linker (yellow). The topology of the receptor-binding domain is not shown, but its position relative to the dscFimH pilin domain is indicated by the labeled box.

The Hultgren group has now proceeded to test whether the provision of the missing strand to pilus subunits obviates the requirement for the chaperone during pilin folding (2). They make use of the mannose-binding FimH lectin assembled into the tip structure of type 1 pili. These pili are expressed by most E. coli strains and have been shown to be required for E. coli to cause bladder infections in mice as well as in primates (15, 16). The FimH lectin consists of an N-terminal lectin domain and a C-terminal pilin domain. They have constructed a FimH variant, donor-strand-complemented FimH (dscFimH), which has been extended at its C terminus by a sequence corresponding to the N terminus of FimG, which is thought to complete the Ig fold of the pilin domain of FimH in the mature pilus (Fig. 1C). The donor strand complementation and exchange model suggests that this donor strand should occupy the groove and complete the fold of the pilin domain of FimH and thus allow it to fold in the absence of the chaperone. Indeed, unlike wild-type FimH, dscFimH can be expressed as a proteolytically stable protein with wild-type mannose-binding activity in the bacterial periplasm in the absence of the periplasmic chaperone FimC. Also in in vitro folding assays, denatured dscFimH, unlike denatured wild-type FimH, was capable of resuming its native β-sheet CD-structure. As predicted by the donor strand exchange model, the dscFimH protein was not incorporated into pili in the presence of chaperone and usher because the groove of the fully folded dscFimH is presumably occupied by the donor strand and thus not free either to interact with the chaperone or to undergo subsequent donor strand exchange. The results indicate that the mechanism of action of a small chaperone like PapD is to provide the missing information (i.e., the missing strand) that allows pilin subunits to fold properly.

Donor strand complementation and exchange also suggest a molecular model for another general role of chaperones in organelle biogenesis. PapD caps subunit interactive surfaces and prevents inappropriate interactions until the subunit has reached its proper assembly site. This is strikingly reminiscent of the invariant chain, which occupies the MHC-class-II peptide-binding groove during folding to prevent nonproductive peptide binding (17). In both of these cases, the interactive groove is first transiently occupied by a chaperone to prevent premature interactions. Subsequently, the chaperone is exchanged for the final substrate, which now permanently occupies the groove. The many chaperones that participate in organelle biogenesis in bacteria and eukaryotic cells may act in an analagous fashion.

The findings presented by the Hultgren group have important practical implications for vaccine development. FimH is a promising vaccine candidate against lower urinary tract infections (16). However, purification of the intact adhesin directly from the polymerized fiber is extremely inefficient and requires the presence of detergents. In addition, FimH cannot be purified when expressed alone because it is proteolytically degraded. Thus, pilus-associated adhesins have been prepared from the periplasmic space in complex with their cognate chaperones and have been used as protective antigens (16). By adding the missing strand to pilus-associated adhesins, it should now be possible to produce proteolytically stable adhesins in large quantities in the absence of chaperone. There is good hope that such structurally “repaired” adhesins could also act as protective antigens because the receptor binding region of known adhesins are located in a domain that is apart from the domain that requires additional steric information.

In summary, the biogenesis of disease-associated bacterial pili has generated fascinating novel insights regarding chaperone-assisted protein folding through donor strand complementation and also has provided an explanation for ordered protein incorporation into heteropolymeric structures.