Substrate-specific binding of hook-associated proteins by FlgN and FliT, putative chaperones for flagellum assembly


Colin Hughes. E-mail; Tel./Fax (+44) 1223 333732.


During flagellum assembly by motile enterobacteria, flagellar axial proteins destined for polymerization into the cell surface structure are thought to be exported through the 25–30 Å flagellum central channel as partially unfolded monomers. How are premature folding and oligomerization in the cytosol prevented? We have shown previously using hyperflagellated Proteus mirabilis and a motile but non-swarming flgN transposon mutant that the apparently cytosolic 16.5 kDa flagellar protein FlgN facilitates efficient flagellum filament assembly. Here, we investigate further whether FlgN, predicted to contain a C-terminal amphipathic helix typical of type III export chaperones, acts as a chaperone for axial proteins. Incubation of soluble radiolabelled FlgN from Salmonella typhimurium with nitrocellulose-immobilized cell lysates of wild-type S. typhimurium and a non-flagellate class 1 flhDC mutant indicated that FlgN binds to flagellar proteins. Identical affinity blot analysis of culture supernatants from the wild-type and flhDC, flgI, flgK, flgL, fliC or fliD flagellar mutants showed that FlgN binds to the flagellar hook-associated proteins (HAPs) FlgK and FlgL. This was confirmed by blotting artificially expressed individual HAPs in Escherichia coli. Analysis of axial proteins secreted into the culture medium by the original P. mirabilis flgN mutant demonstrated that export of FlgK and FlgL was specifically reduced, with concomitant increased release of unpolymerized flagellin (FliC), the immediately distal component of the flagellum. These data suggest that FlgN functions as an export chaperone for FlgK and FlgL. Parallel experiments showed that FliT, a similarly small (14 kDa), potentially helical flagellar protein, binds specifically to the flagellar filament cap protein, FliD (HAP2), indicating that it too might be an export chaperone. Flagellar axial proteins all contain amphipathic helices at their termini. Removal of the HAP C-terminal helical domains abolished binding by FlgN and FliT in each case, and polypeptides comprising each of the HAP C-termini were specifically bound by FlgN and FliT. We suggest that FlgN and FliT are substrate-specific flagellar chaperones that prevent oligomerization of the HAPs by binding to their helical domains before export.


Enterobacterial flagella are long cell surface appendages essential for motility, comprising three distinct contiguous substructures: a basal body anchored in the cell envelope; a flexible hook extending from the cell surface; and a long helical filament that acts as a propeller (Macnab, 1996). Over 40 genes are required for the construction of a functional flagellum, with 18 encoding flagellar structural components (Aizawa, 1996). These can be divided into two classes: those that polymerize to form rings that anchor the flagellum in the cell envelope; and those that polymerize to form the hollow filamentous axial structure, up to 15 μm long, that traverses both the cytoplasmic and the outer membranes (Fig. 1). The flagellar axial proteins that constitute the basal body rod, the hook, the hook–filament junction, the filament and the filament cap are exported via the flagellum-specific secretion pathway through the 25–30 Å central channel in the flagellum (Homma et al., 1990; Ruiz et al., 1993). Because of the small diameter of the export channel, it is thought that the axial subunits are exported as partially unfolded monomers, but the mechanisms preventing folding and oligomerization of the axial proteins in the cytosol before export are not known (Namba and Vonderviszt, 1997). Assembly of the axial structure proceeds from base to tip, each subunit being added to the distal end of the growing flagellum (Iino, 1969; Emerson et al., 1970; Ohnishi et al., 1994). The order in which proteins are exported is partly controlled at the level of transcription, the components of the basal body and the hook being expressed and exported before transcription of the genes encoding the hook-associated proteins (HAPs; FlgK, FlgL and the cap protein FliD) and the filament protein (FliC) (Kutsukake et al., 1990; Macnab, 1992; Liu and Matsumura, 1994). Transcription of these late, or class 3, flagellar genes is coupled to the morphogenic status of the flagellum via the anti-sigma factor FlgM (Ohnishi et al., 1992; Hughes et al., 1993; Kutsukake and Iino, 1994).

Figure 1.

. Ring and axial proteins of the enterobacterial flagellum. om, outer membrane; cm, cytoplasmic membrane.

While control of flagellar gene transcription prevents unnecessary expression of FliC and the other cell-distal components of the flagellum, little is known about the mechanisms that govern the ordered assembly of the axial proteins. The initiation of axial substructure assembly is sequential, i.e. the rod must be completed before the initiation of hook assembly, and a complete hook is required before the HAPs can assemble (Kubori et al., 1992). The one exception to this is assembly of the filament cap, thought to be a FliD pentamer (Maki et al., 1998), which must be assembled upon completion of the FlgK/FlgL hook–filament junction to allow polymerization of the FliC filament to proceed (Homma et al., 1984a,b; Ikedaet al., 1993). As the filament assembles, FliD is continually displaced to the end of the growing filament, where it acts as a nucleation point for FliC polymerization (Ikeda et al., 1985; 1987). In a fliC mutant, FliD remains associated with FlgL (HAP3) and FlgK (HAP1) at the tip of the hook and, for this reason, it was originally classified as HAP2 (Homma et al., 1984a).

Earlier studies in Salmonella typhimurium and in the hyperflagellated swarming Proteus mirabilis indicated that the flagellar protein FlgN facilitates the initiation of filament assembly (Kutsukake et al., 1994; Gygi et al., 1997). A motile but non-swarming P. mirabilis flgN mutant exhibited a leaky defect in filament assembly characterized by a severe reduction in FliC incorporation into membrane-anchored flagellar filaments, concomitant with the increased secretion of unpolymerized FliC into culture supernatants (Gygi et al., 1997). Loss of FlgN in S. typhimurium and P. mirabilis results in a reduction in the number of flagella possessing filaments (Kutsukake et al., 1994; Gygi et al., 1997). The flgN mutant phenotype closely resembles the flagellar assembly defects of S. typhimurium mutants lacking HAPs (Homma et al., 1984b), and we have suggested a possible role for FlgN as a cytosolic chaperone of one or more of the HAP proteins. This suggestion is strengthened by the predicted structural similarity of FlgN to type III secretion chaperones (Gygi et al., 1997), such as SycE and SycH of Yersinia (Wattiau et al., 1994; 1996). Like FlgN, each of these virulence chaperones is small with a putative C-terminal amphipathic helix, proposed to mediate interactions with target secreted proteins (Woestyn et al., 1996; Hueck, 1998). We observed these predicted characteristics not only in FlgN, but also in FliT, a flagellar protein of unascribed function encoded in an operon with FliD (Yokoseki et al., 1995), the loss of which abolishes motility in Escherichia coli (Kawagishi et al., 1992). In this report, we provide evidence that FlgN and also FliT act as specific HAP chaperones.


The flagellar proteins FlgN and FliT, possible export chaperones

FlgN and FliT have calculated molecular weights (MW) of 16.5 kDa and 14 kDa respectively. They do not show significant primary sequence similarity to each other or to other proteins currently in the databases, but their C-terminal regions both contain heptad repeats of hydrophobic residues that are predicted to form amphipathic helices (Fig. 2), similar to those in the C-termini of substrate-specific chaperones required for efficient type III export of virulence proteins, e.g. SycE of Yersinia enterocolitica (Wattiau et al., 1996). FlgN also has an N-terminal leucine zipper-like motif (Fig. 2) that has a high probability (> 0.99) of coiled-coil formation, as determined by the algorithm of Lupas et al. (1991).

Figure 2.

. Potential amphipathic helices in FlgN and FliT. α-Helical wheel projections of S. typhimurium FlgN N-terminal residues 1–30, FlgN C-terminal residues 74–114 and FliT C-terminal residues 60–98. Hydrophobic residues are circled; charged residues shown as ±.

The wild-type S. typhimurium flgN and fliT genes were amplified by polymerase chain reaction (PCR) using the primer pairs N1 and N2, or T1 and T2 respectively (Table 1). The genes were inserted separately into the T7 expression vector pET11c (Studier et al., 1990), creating the plasmids pETflgN and pETfliT (Table 2). SDS–PAGE (15%) analysis of [35S]-methionine whole-cell lysates of IPTG-induced E. coli BL21 (DE3) carrying either pETflgN or pETfliT confirmed the synthesis of FlgN and FliT proteins of approximately 16 kDa and 14 kDa, respectively, and indicated that 90% of the FlgN produced was soluble, compared with 5% of the FliT (data not shown).

Table 1. . Oligonucleotides.Thumbnail image of
Table 2. . Plasmids. a. Studier et al. (1990).b. Invitrogen, all other plasmids this work.Thumbnail image of

FlgN and FliT specifically bind to flagellar proteins in bacterial cell lysates

Affinity blotting of putative chaperones to immobilized substrates has been used to demonstrate in vitro binding of type III secretion chaperones to secreted virulence factors (Wattiau and Cornelis, 1993; Woestyn et al., 1996; Wainwright and Kaper, 1998). To determine whether FlgN and FliT interact in a similar way with flagellar proteins, the putative chaperones were affinity blotted to whole-cell lysates of wild-type S. typhimurium SJW1103 and the derivative flhDC mutant strain SJW1368, which lacks the FlhDC transcriptional regulator of the flagellar gene hierarchy and, therefore, does not produce any flagellar proteins. Lysates of both strains were separated by SDS–PAGE (15%) and electrotransferred to nitrocellulose. These blots were incubated with crude soluble extracts of E. coli BL21 (DE3) carrying the T7 expression plasmids pETflgN or pETfliT and expressing either FlgN or FliT, respectively, radiolabelled with [35S]-methionine. Subsequent autoradiography of the filters revealed that FlgN and FliT bound to proteins in the wild-type cell lysate but did not bind to proteins in the lysate of the flhDC strain (Fig. 3). Radiolabelled FlgN bound to two proteins of approximately 60 kDa and 34 kDa in the wild-type lysate, whereas FliT bound to one protein of approximately 50 kDa.

Figure 3.

. Affinity blot of FlgN and FliT to whole-cell lysates of S. typhimurium. Late-exponential cultures of S. typhimurium SJW1103 (WT) and the derived flhDC mutant SJW1368 (flhDC) were centrifuged, and cells were lysed in urea–SDS loading buffer. Proteins (from 1 ml of cells; A600 of 0.05) were separated by SDS–PAGE (15%), stained with Coomassie blue (left) and probed with [35S]-methionine FlgN (centre) or FliT (right). Molecular weight markers (New England Biolabs prestained protein marker; broad range) are in kDa.

FlgN and FliT each bind to specific flagellar axial proteins

A subset of the flagellar axial proteins is typically exported in excess and released by wild-type cells into the extracellular medium, i.e. the hook protein (FlgE) and the HAPs (FlgK, FlgL and the filament cap FliD) (Homma and Iino, 1985a). In addition, sheared FliC filaments are present in wild-type culture supernatants (Homma and Iino, 1985a), and secreted FliC monomers (MW 51.2 kDa) are found in the culture supernatants of mutants specifically lacking FlgK, FlgL or FliD (Homma et al., 1984b). Culture supernatants can thus be collected and concentrated to provide an enriched source of these exported flagellar axial proteins. To determine whether FlgN and FliT bind to one or more of these extracellular proteins, culture supernatants were collected from wild-type S. typhimurium and from flagellar mutants containing specific disruptions in flgK, flgL, fliC, fliD, flhDC (abolishing the production of flagellar proteins) or flgI (preventing the export of cell-distal axial proteins). Coomassie staining of the precipitated S. typhimurium culture supernatants detected a number of proteins secreted by all of the wild-type and flagellar mutant strains, but identified several proteins secreted by the wild type and the class 3 flagellar mutants (flgK, flgL, fliC or fliD) but not the flhDC or flgI mutants (Fig. 4A).

Figure 4.

. FlgN and FliT affinity blots of S. typhimurium culture supernatants. A. Secreted proteins in culture supernatants of S. typhimurium SJW1103 (WT) and derived flagellar mutants, SJW1368 (flhDC), SJW1351 (flgI ), SJW2177 (flgK ), SJW2172 (flgL), SJW2536 (fliC) and SJW2149 (fliD) were separated by SDS–PAGE (15%) and, in parallel, stained with Coomassie blue (left) or affinity blotted with [35S]-methionine FlgN (centre) and FliT (right). B. Parallel immunoblots with anti-FlgK (top), anti-FliD (middle) and anti-FlgL (bottom) antisera. The position of flagellar hook-associated protein bands detected by antibody correlates with the protein bands bound by FlgN (corresponding to FlgK and FlgL) and FliT (corresponding to FliD) observed in (A). Molecular weight markers are in kDa.

Affinity blot analysis showed that radiolabelled FlgN and FliT bound proteins in the culture supernatants (Fig. 4A). In both wild-type and fliC mutant culture supernatants, FlgN bound to proteins of 60 kDa and 34 kDa, whereas FliT bound to a protein of 50 kDa, as observed on the previous whole-cell affinity blots. There was no binding to proteins secreted by the flhDC (class 1) or flgI (class 2) mutants, indicating that FlgN and FliT recognized exported flagellar proteins, possibly encoded by class 3 genes. Affinity blotting of proteins secreted by the flgK, flgL and fliD mutant strains suggested that FlgN and FliT bound to proteins encoded by these genes. FlgN bound to two proteins secreted by the fliD mutant but, of the proteins secreted by the flgK mutant, only the 34 kDa protein was bound and, of the proteins secreted by the flgL mutant, only the 60 kDa protein was bound. This indicates that FlgN binds to both the 60 kDa FlgK and the 34 kDa FlgL. FliT bound to a 50 kDa protein secreted by the flgK and flgL mutants but did not bind to any of the proteins secreted by the fliD mutant, suggesting that FliT binds specifically to the 50 kDa FliD. Neither FlgN nor FliT bound to any other supernatant protein, including FliC, which was the most abundant protein in the wild-type and mutant supernatants.

The above interpretation was strengthened by parallel immunoblots of the culture supernatants with anti-FlgK, anti-FlgL and anti-FliD antisera. These confirmed that FlgK, FlgL and FliD were secreted by both the wild-type and fliC mutant but were not secreted into the flhDC and flgI mutant supernatants (Fig. 4B). As expected, strains mutated in flgK, flgL or fliD lost only the corresponding protein from their culture supernatants, e.g. the fliD mutant did not secrete FliD but did secrete FlgK and FlgL.

FlgN and FliT bind to C-terminal helical domains of specific artificially expressed HAPs

To confirm that FlgN and FliT bind to specific HAPs, the flgK, flgL and fliD genes were placed under the control of the expression vector pET11c T7 promoter, creating the plasmids pETflgK, pETflgL and pETfliD (Table 1). The HAPs were expressed in E. coli BL21 (DE3), and cell lysates were analysed by SDS–PAGE (15%) and Coomassie staining (Fig. 5A). The apparent sizes of the overexpressed proteins were consistent with their calculated molecular weights, FliD (466 amino acids) being 50 kDa, FlgK (552 amino acids) 60 kDa and FlgL (317 amino acids) 34 kDa. Affinity blots of whole cells expressing the HAPs confirmed that FlgN bound to FlgK and FlgL, and FliT bound to FliD. The binding of radiolabelled FlgN and FliT was specific, as there was no cross-reactivity between FlgN and FliD, or FliT and FlgK or FlgL (Fig. 5A). Binding of radiolabelled FlgN or FliT could be quenched approximately five- to 10-fold by incubation of affinity blots in the presence of cell lysates containing 100-fold excess of unlabelled FlgN or FliT (Fig. 5B). We have no information on the nature of the apparent fliD doublet observed here and less obviously in Fig. 4.

Figure 5.

. Specific binding of FlgN and FliT to overexpressed flagellar HAPs. A. Late-exponential cultures of IPTG-induced E. coli BL21(DE3) carrying the T7 expression vector pET11c (−) or recombinant derivatives pETfliD (FliD), pETflgK (FlgK) and pETflgL (FlgL) were centrifuged, and cells were lysed in urea–SDS loading buffer. Proteins (from 1 ml of cells; A600 of 0.05) were separated by SDS–PAGE (15%) and either stained with Coomassie brilliant blue (left) or affinity blotted with [35S]-methionine FlgN (centre) or FliT (right). Molecular weight markers are in kDa. B. IPTG-induced E. coli BL21 (DE3) carrying pETK (FlgK), pETL (FlgL) and pETD (FliD) were collected and lysed as described above. Proteins (from 1 ml of cells; A600 of 0.01, 0.02, 0.05 and 0.10 from left to right) were separated by SDS–PAGE (15%) and affinity blotted with either [35S]-methionine FlgN (top) or FliT (bottom) in the absence (−) or presence (+) of an approximately 100-fold excess of unlabelled FlgN (top) or FliT (bottom).

The N- and C-terminal regions of FlgK and FlgL and the C-terminal region of FliD are devoid of prolines and contain heptad repeats of hydrophobic residues (Homma et al., 1990). These terminal domains potentially form amphipathic helices (Vonderviszt et al., 1992), similar to those predicted in FlgN and FliT (Fig. 2). N- and C-terminal truncates of the FlgK, FlgL and FliD proteins (Figs 6A and 7A) were constructed by inserting flgK, flgL, fliD and deleted derivatives of these genes, amplified by PCR, into the T7 expression vector pET15b (Table 1). SDS–PAGE (15%) analysis of cell lysates of E. coli BL21 (DE3) overexpressing FlgK, FlgL and their respective truncates confirmed the production of stable polypeptides of the expected sizes (Fig. 6B, left). Affinity blot analysis of these lysates showed that binding of radiolabelled FlgN was not diminished by removal of the N-terminal helices of FlgK (producing a 56 kDa polypeptide lacking amino acids 1–26) or FlgL (producing a 29 kDa polypeptide lacking amino acids 1–29), but polypeptides composed of the N-terminal 99 amino acids of FlgK and the N-terminal 76 amino acids of FlgL were no longer bound by FlgN. Removal of the C-terminal 42 amino acids of the 552 residue FlgK and the C-terminal 55 amino acids of the 317 residue FlgL completely abolished binding of FlgN (Fig. 6B, right). Moreover, polypeptides comprising the C-terminal 69 residues of FlgK (8 kDa) and the C-terminal 67 residues of FlgL (7 kDa) were bound by FlgN, albeit to a lesser degree (Fig. 6B), suggesting that these short C-terminal helices of FlgK and FlgL are central to the binding of FlgN.

Figure 6.

. Affinity blotting of FlgK and FlgL truncates with FlgN. A. Truncated FlgK and FlgL, encoded by deleted derivatives of pETK and pETL (Table 1) lacking predicted N- and C-terminal amphipathic helices (shown as shaded boxes). B. E. coli BL21 (DE3) cells artificially overexpressing either FlgK, truncated derivatives of FlgK (Δ1–26, N99, Δ510–552 and C69), FlgL or truncated derivatives of FlgL (Δ1–29, N76, Δ262–317 and C67) were lysed in urea–SDS loading buffer. Proteins (from 1 ml of cells; A600 of 0.05) were separated by SDS–PAGE (15%) and either stained with Coomassie blue (left) or, in parallel, affinity blotted with [35S]-methionine FlgN (right).

Figure 7.

. Affinity blotting of truncated FliD with FliT. A. Truncated FliD, encoded by deleted derivatives of pETD (Table 1), lacking predicted C-terminal amphipathic helix (shown as shaded box). B. E. coli BL21 (DE3) artificially overexpressing either FliD or its truncated derivatives (Δ428–466 and C127) were lysed in urea–SDS loading buffer. Proteins (from 1 ml of cells; A600 of 0.05) were separated by SDS–PAGE (15%) and stained with Coomassie blue (left) or affinity blotted with [35S]-methionine FliT (right). Molecular weight markers are in kDa.

N- and C-terminal truncates of FliD (Fig. 7A) were analysed similarly to determine their affinity for radiolabelled FliT (Fig. 7B). Removal of the 39-amino-acid C-terminal helical domain of the 466 residue FliD, producing a 427-amino-acid polypeptide of 47 kDa, abolished the binding of radiolabelled FliT. In contrast to this, a 14 kDa polypeptide containing only the C-terminal 127 residues of FliD was bound by FliT, suggesting that, like the binding of FlgN to FlgK and FlgL, the C-terminal predicted amphipathic helix of FliD is the domain that mediates binding of FliT.

A flgN mutant is specifically defective in the export of FlgK and FlgL

If FlgN acts as a substrate-specific chaperone facilitating efficient secretion of FlgK and FlgL, a flgN mutant should be specifically deficient in the export of these target axial proteins. To assess this, the secretion of axial proteins was determined in the original P. mirabilis flgN transposon mutant (Gygi et al., 1997). Cells from mid-exponential cultures of wild-type P. mirabilis and a derived daughter flgN mutant were harvested, resuspended in fresh LB medium and incubated for a further 15 min, before the collection and precipitation of newly secreted proteins. In addition to giving a clearer view of export without excessive extracellular accumulation, this procedure minimized the degradation of secreted axial proteins of P. mirabilis, which, unlike flagellar proteins of S. typhimurium, are only transiently stable in the extracellular medium (J. C. Q. Bennett, unpublished data). As controls, further isogenic flagellar mutants with specific disruptions in flhDC (class I) and flhA (class II) genes were assayed in parallel.

Coomassie staining of proteins isolated from wild-type supernatants revealed an approximately 39 kDa protein, which was subsequently confirmed to be FliC by Western blotting (data not shown), and three other proteins of 58, 56 and 34 kDa that were not detectable in the supernatants of the fhDC or flhA mutants (Fig. 8). Affinity blotting of these proteins using radiolabelled P. mirabilis FlgN, expressed in E. coli BL21 (DE3) from the T7 promoter of the recombinant plasmid pETflgN2, indicated that the 58 and 34 kDa proteins corresponded to FlgK and FlgL respectively. As shown previously, supernatant from the flgN mutant showed an approximately 30-fold increase in the amount of extracellular unpolymerized FliC compared with wild type (Gygi et al., 1997) and, in addition, secretion of the 56 kDa putative FliD protein was enhanced approximately 10-fold. Concomitant with these increases, affinity blot analysis and phosphorimager scanning revealed approximately 15- and twofold reductions in the amount of FlgK and FlgL, respectively, in the supernatant of the flgN mutant. These data, obtained in each of three separate experiments, confirm that FlgN specifically affects the secretion of both FlgK and FlgL, with a greater apparent effect on FlgK.

Figure 8.

. Affinity blot of P. mirabilis culture supernatants with FlgN. Secreted proteins in culture supernatants of P. mirabilis U6450 (WT) and derived flagellar mutants, flhDC, flhA and flgN were separated by SDS–PAGE (12.5%) and stained with Coomassie blue (left), or, in parallel, affinity blotted with [35S]-methionine P. mirabilis FlgN (right). Molecular weight markers are in kDa.


The bacterial flagellum-specific export apparatus is referred to as the archetypal type III secretion system (Stephens and Shapiro, 1996). While a large number of putative components involved in flagellar synthesis and assembly have been identified, less is known about the mechanisms underlying the export of flagellar axial proteins through the hollow core of the flagellum to their sites of oligomerization (Macnab, 1996). In type III virulence protein secretion systems, e.g. Yop secretion of Yersinia, cytosolic substrate-specific chaperones maintain secreted proteins in an export-competent conformation (Wattiau et al., 1996). Here, we provide evidence that two small flagellar proteins, FlgN and FliT, perform an analogous function during flagellar export.

To assess the potential chaperone behaviour of FlgN and FliT, we used the method of affinity blotting in which proteins separated by SDS–PAGE are transferred to nitrocellulose filters that are subsequently incubated with the soluble radiolabelled putative chaperone. Blotting of radiolabelled FlgN or FliT to wild-type S. typhimurium whole-cell lysates and cell-free culture supernatants, which contain large quantities of secreted flagellar components (Homma and Iino, 1985, Homma et al. 1984b; Hughes et al., 1993), indicated that FlgN binds to proteins of 60 kDa and 34 kDa, and FliT binds to a 50 kDa protein. Neither FlgN nor FliT recognized any proteins in the cell lysate or culture supernatant of a mutant lacking the class 1 flhDC flagellar master operon, the expression of which is a prerequisite for the synthesis of all class 2 and class 3 flagellar proteins (Bartlett et al., 1988; Liu and Matsumura, 1994; Furness et al., 1997). Similarly, FlgN and FliT did not bind to any of the proteins secreted into the culture supernatant by the flgI mutant, which lacks the flagellar P-ring and, consequently, cannot export or assemble the cell-distal components of the flagellum, i.e. FlgE, FlgK, FlgL, FliC and FliD (Kubori et al., 1992). These data suggest that FlgN and FliT bind to specific axial components of the flagellum that are exported beyond the outer membrane.

Affinity blotting of proteins in the culture supernatants of mutants specifically lacking FliC or the HAPs (FlgK, FlgL and FliD) indicated that FlgN binds to the 60 kDa FlgK and the 34 kDa FlgL, and that FliT binds to the 50 kDa FliD. This was confirmed by affinity blotting cell lysates of E. coli artificially expressing each of the HAPs from recombinant expression plasmids. The interactions of FlgN with FlgK and FlgL, and FliT with FliD were specific, as there was no cross-reactivity of the putative chaperones with other proteins, and the binding of each radiolabelled chaperone to its target HAP could be quenched by the incubation of affinity blots with unlabelled cognate chaperone. Neither FlgN nor FliT was visible in culture supernatants containing large quantities of HAPs, suggesting that they are not co-exported with the HAPs. It is possible that other flagellar axial proteins, i.e. FliC and the hook protein FlgE, are chaperoned before export, but our data strongly suggest that this is not performed by FlgN and FliT. It is noteworthy that a S. typhimurium fliS mutant produces short flagella and accumulates FliC in the cytosol (Yokoseki et al., 1995), suggesting that the 15 kDa FliS protein may be required for the efficient export of FliC. While the predicted secondary structure of FliS is highly helical, unlike FlgN and FliT, it does not possess a C-terminal amphipathic helix. Attempts to demonstrate in vitro interaction of FliS with FliC are equivocal in our hands (G. M. Fraser and J. C. Q. Bennett, unpublished).

The P. mirabilis flgN mutant showed altered stoichiometry of secreted axial proteins. The 15-fold and twofold reduction in the levels of extracellular FlgK and FlgL detected by affinity blotting caused a 30-fold increase in flagellin monomers secreted into the culture supernatant, as inefficient HAP secretion hinders FliC polymerization into an extending filament. While the flgN mutant data appear to suggest that FlgN has a stronger effect on FlgK than its immediately distal HAP component FlgL, prompting ideas of sequential component selectivity, it is likely that the decrease in FlgL may be underestimated from our data. This is likely because, in the absence of completed FlgK rings, any exported FlgL will tend to leak into the supernatant rather than be incorporated into the flagellum. It is possible that FlgL export is as impaired as that of FlgK.

A common characteristic of the flagellar axial proteins is the presence of amphipathic helices at their N- and C-termini (Homma et al., 1990). In axial protein monomers, the terminal domains are solvent exposed and have a random coil structure but, upon polymerization, the coils form ordered bundles that are thought to play a role in quaternary interactions between subunits, maintaining the overall stability and conformation of the axial substructures (Kostyukova et al., 1988; Vonderviszt et al., 1989; 1991; 1995). Neither FlgN nor FliT bound to truncates of the HAPs lacking the C-terminal domains, whereas polypeptides spanning the C-terminal helices of the HAPs were able to bind their cognate chaperone.

That FlgN and FliT appear to recognize a specific domain on their target proteins contrasts with the cytosolic ‘housekeeping’ chaperones, e.g. DnaK and GroEL, which bind to multiple hydrophobic sites on a wide range of proteins (Netzer and Hartl, 1998). It is reminiscent of Yops recognition by the specific Yersinia Syc chaperones that bind to an approximately 35–70 amino acid N-terminal domain that is required for the translocation of Yops across eukaryotic membranes (Sory et al., 1995; Woestyn et al., 1996). The N-terminal, rather than the C-terminal, location of the Yop chaperone-binding domain may reflect the functional differences between the Yop virulence proteins and the flagellar axial proteins. The Syc chaperones are proposed to be anti-aggregation factors (Woestyn et al., 1996), but may also be involved in guiding secreted substrates to the type III export apparatus (Cheng et al., 1997; Lee et al., 1998) in much the same way as SecB functions as both an anti-folding factor and a secretion pilot for proteins exported via the signal peptide-dependent general secretion pathway (Hartl et al., 1990; Driessen et al., 1998). Given that flgN and fliT mutants are able to construct functional, albeit fewer, flagella (Gygi et al., 1997; Yokoseki et al., 1995), it is unlikely that the flagellar chaperones act as secretion pilots. Nevertheless, it is possible that, as the machinery mediating type III secretion of virulence proteins is thought to have evolved from the flagellar export apparatus, the Syc chaperones developed from their flagellar counterparts.

Why do flagellar components require secretion chaperones? The axial components of the flagellum are capable of oligomerization in vitro. Both FlgE, the structural subunit of the hook, and FliD can self-assemble (Kato et al., 1982; Ikeda et al., 1996), and FliC can polymerize to form filaments if provided with a nucleation point (Asakura et al., 1964) or in the presence of a precipitant such as ammonium sulphate (Ada et al., 1963). Despite the propensity of the axial proteins to form oligomers, they are thought to exist as monomers in the cell cytosol and to be exported in a partially unfolded monomeric state (Namba and Vonderviszt, 1997). This suggests that either conditions in the cytosol are unfavourable for polymerization of flagellar components or there are factors in the cytosol that maintain flagellar proteins in a monomeric form. It is plausible that FlgN and FliT act as ‘bodyguards’, protecting the exposed unfolded C-terminal domains of the HAPs. Like the periplasmic chaperone PapD, which is specifically required for the export and assembly of P pilus structural proteins (Thanassi et al., 1998), FlgN and FliT may prevent inappropriate interactions of the HAPs before secretion by binding their interactive surfaces (Fig. 9). Upon recognition of the HAP–chaperone complex by the flagellar export apparatus, the chaperone might be released, allowing further cycles of binding to nascent HAP monomers.

Figure 9.

. Model for FlgN and FliT chaperone activity. om, outer membrane; cm, cytoplasmic membrane; shaded boxes/circles, N-/C-terminal amphipathic helices; unshaded box, FliD N-terminal domain.

Experimental procedures

Bacterial strains and recombinant DNA techniques

S. typhimurium SJW1103 has wild-type motility (Yamaguchi et al., 1994), and its derivatives are mutated in flagellar genes, i.e. SJW1368 (flhDC), SJW1351 (flgI ), SJW2172 (flgL), SJW2177 (flgK ), SJW2149 (fliD), SJW2536 (fliC) (Ohnishi et al., 1994). The P. mirabilis motile, non-swarming flgN mutant (Gygi et al., 1997) and non-motile, non-swarming flhDC mutant (R. Furness, unpublished) are derivatives of wild-type P. mirabilis U6450 (Allison and Hughes, 1991). Bacteria were grown at 37°C in Luria–Bertani (LB) broth or on LB agar, supplemented with ampicillin (100 μg ml−1), kanamycin (50 μg ml−1) and spectinomycin (50 μg ml−1) when necessary. Routine DNA manipulation and electroporation were carried out as described previously (Sambrook et al., 1989) using E. coli recA1 XL1 Blue (Stratagene). Oligonucleotides and plasmids are listed in Tables 1 and 2. S. typhimurium fliD, flgK, flgL, flgN and fliT genes and P. mirabilis flgN were amplified by PCR using either S. typhimurium SJW1103 or P. mirabilis U6450 chromosomal DNA template (1–10 ng) in native Pfu buffer (Stratagene), 0.25 mM each dNTP and 50 pmol of the corresponding oligonucleotide pairs, with 2.5 U of native Pfu DNA polymerase (Stratagene) in a Perkin-Elmer thermal cycler. Amplified DNA was purified using Wizard PCR Preps (Promega), digested with NdeI and BamHI and ligated to NdeI/BamHI-digested T7 expression vectors pET11c and pET15b (Studier et al., 1990; Invitrogen).

Assay of extracellular flagellar proteins

S. typhimurium cultures (50 ml) were grown to late-exponential phase (A600 of 1.0), centrifuged at 10 000 g and culture supernatants collected and centrifuged twice further to remove cells and cell debris. Supernatants were filtered through a 0.45 μm filter (Millipore). TCA was added to 10% (w/v), and proteins were precipitated on ice for 1 h. Proteins were separated by SDS–PAGE (15%) (Laemmli, 1970) and stained with Coomassie brilliant blue. Immunoblotting was carried out as described previously (Sambrook et al., 1989), after separated proteins were transferred to nitrocellulose (Hybond-C; Amersham). Anti-FlgL and anti-FlgK antisera were used at a concentration of 1:10 000 and anti-FliD antisera at a concentration of 1:2500 (antisera were a gift from Takeshi Ikeda, Nagoya University). After incubation of blots with goat anti-rabbit IgG–horseradish peroxidase (HRP) conjugate (Pierce), chemiluminescent detection was carried out using Supersignal substrate (Pierce).

Artificial expression of flagellar proteins

E. coli BL21 (DE3) carrying the appropriate plasmid was grown to mid-exponential phase (A600 of 0.7), and T7-controlled expression of protein was induced by the addition of IPTG to 1 mM (Studier and Moffatt, 1986). Cultures were incubated for a further 3 h, after which cells (A600 of 1.0) were spun down and resuspended in 1 ml of 8 M urea SDS–PAGE loading buffer. Expression of radiolabelled FlgN or FliT in E. coli BL21 (DE3) carrying either pETflgN or pETfliT was induced in the presence of rifampicin (5 μg ml−1), and proteins were labelled with [35S]-methionine (90 mCi; Amersham). Labelled cells were fractionated (Koronakis et al., 1991), and soluble and insoluble fractions were separated by centrifugation at 16 000 g for 10 min.

Affinity blotting with FlgN and FliT

Proteins separated by SDS–PAGE (15%) were transferred to nitrocellulose, and membranes were preincubated for 2 h at room temperature in PBS containing 5% w/v skimmed milk. Soluble extracts of E. coli BL21 (DE3) expressing radiolabelled FlgN, FliT or β-lactamase were added directly to the preincubation solution. Competition assays were performed by preincubating and incubating membranes as described in the text with the addition of soluble extracts of E. coli BL21 (DE3) overexpressing unlabelled FlgN, FliT or β-lactamase. Membranes were incubated for 16 h at room temperature, washed three times for 15 min in PBS and dried. Binding of radiolabelled probe was detected by autoradiography.

Assay of flagellar protein export by P. mirabilis

Overnight cultures of P. mirabilis U6450 and derived flagellar mutants flhDC (R. B. Furness, unpublished), flhA (Gygi et al., 1995) and flgN (Gygi et al., 1997) were diluted 1:100 in LB medium and grown at 37°C with shaking to A600 of 0.8. Cells from 50 ml of culture were harvested by centrifugation (13 000 g for 2 min) and resuspended in an equal volume of prewarmed LB. After a further 15 min incubation, secreted proteins in the culture supernatant were prepared as above. BSA (20 μg) was added to each precipitation and used as a standard in subsequent SDS–PAGE (12.5%) analysis to ensure equivalent recovery of proteins (data not shown). Each preparation was repeated in triplicate. Binding of radiolabelled P. mirabilis flgN to affinity blots was detected using a Cyclone phosphorimager (Packard) and quantified using optiquant software (Packard).


  1. †These authors contributed equally to this study


We are very grateful to Robert Macnab (Yale University) and Richard Furness (our laboratory) for providing bacterial strains, and Takeshi Ikeda (Aichi-Gakuin University) and Michio Homma (Nagoya University) for giving us antibodies. We thank Daniel Gygi for useful discussions. The work was funded by a Wellcome Trust Programme grant (C.H.) and Medical Research Council and Wellcome Trust studentships to G.M.F. and J.C.Q.B. respectively.