An escort mechanism for cycling of exportchaperones during flagellum assemblyLewis D. B. Evans, Graham P. Stafford, Sangita Ahmed, Gillian M. Fraser, and Colin Hughes*

Department of Pathology, Cambridge University, Tennis Court Road, Cambridge CB2 1QP, United Kingdom

Edited by Olaf Schneewind, University of Chicago, Chicago, IL, and accepted by the Editorial Board September 26, 2006 (received for review June 21, 2006)

Assembly of the bacterial flagellar filament requires a type IIIexport pathway for ordered delivery of structural subunits fromthe cytosol to the cell surface. This is facilitated by transientinteraction with chaperones that protect subunits and pilot themto dock at the membrane export ATPase complex. We reveal thatthe essential export protein FliJ has a novel chaperone escortfunction in the pathway, specifically recruiting unladen chaper-ones for the minor filament-class subunits of the filament cap andhook-filament junction substructures. FliJ did not recognize un-chaperoned subunits or chaperone-subunit complexes, and it as-sociated with the membrane ATPase complex, suggesting a func-tion postdocking. Empty chaperones that were recruited by FliJ invitro were efficiently captured from FliJ-chaperone complexes bycognate subunits. FliJ and subunit bound to the same region on thetarget chaperone, but the cognate subunit had a �700-fold greateraffinity for chaperone than did FliJ. The data show that FliJ recruitschaperones and transfers them to subunits, and indicate that thisis driven by competition for a common binding site. This escortmechanism provides a means by which free export chaperones canbe cycled after subunit release, establishing a new facet of thesecretion process. As FliJ does not escort the chaperone for themajor filament subunit, cycling may offer a mechanism for exportselectivity and thus promote assembly of the junction and capsubstructures required for initiation of flagellin polymerization.

protein secretion � secretion pilots � type III export

Bacterial motility is commonly conferred by cell surfaceflagella. The long flagellar filament acts as a rotating helical

propeller, and it is anchored to the basal body in the cell envelopevia a flexible hook (1). The filament structure comprises fourtypes of filament-class subunits. The major filament substructurecomprises about 20,000 flagellin (FliC) subunits, which arepolymerized under the distal filament cap, a FliD pentamer thatis displaced farther from the cell as the filament grows. Thefilament is adapted to the flexible hook by a preformed hook-filament junction made up of 11 subunits each of FlgK and FlgL.The filament structure is assembled in strict sequence; the capand junction substructures must be assembled before flagellin isaccepted by the nascent structure.

The major subunit FliC and the minor subunits FliD, FlgK,and FlgL are delivered from the cytosol to the base of the nascentflagellum by a type III export pathway in which they are boundby the subunit-specific export chaperones FliS (for FliC), FliT(for FliD), and FlgN (for FlgK and FlgL). These export chap-erones effect transition to the membrane by preventing prema-ture polymerization of subunits, acting as ‘‘bodyguards’’ for theC-terminal amphipathic oligomerization domain (2–4) and bypiloting the subunits to the export apparatus (5). The chaper-oned subunits dock at the membrane-associated export ATPaseFliI (5, 6), a hexameric ring with a central pore proposed to alignwith the central channel of the nascent flagellum (7–9). Tran-sient intermediate complexes underlying the early stages of theexport pathway have been identified, but there are few dataabout the series of events postdocking. In vitro studies of arelated type III pathway have shown that ATP hydrolysis by theexport ATPase facilitates chaperone release (10). However,

nothing is known about the fate of chaperones once releasedfrom the filament-class subunits, and there is no evidence forsubunit selectivity by the export apparatus despite the stringentdelivery sequence and striking subunit stoichiometry in thecompleted flagellum.

To study further the sequence of events underlying the exportpathway we have addressed the function of FliJ, which is essentialfor export of structural subunits (11). This unveiled an unpre-dicted and novel activity indicating cycling of export chaperones.

ResultsFliJ Is Not an Export Chaperone for Hook- or Filament-Class StructuralSubunits. FliJ resembles subunit chaperones, e.g., in size, and hasbeen suggested to act as a ‘‘general’’ cytosolic chaperone (12).We sought to isolate the predicted transient complexes of FliJand subunits by performing affinity chromatography of E. coliextracts containing overexpressed untagged hook-and-filament-class subunits, using (His6)-tagged FliJ. This FliJ variant is activeas it complements a fliJ null mutant. None of the flagellarhook-class subunits FlgD, FlgE, and FliK or filament-classsubunits FlgM, FlgK, FlgL, FliD, and FliC was bound by(His6)FliJ, assayed following Ni2�-affinity copurification (Fig. 1Upper). As has been reported by other laboratories (13, 14),(His6)FliJ did bind to FliH (Fig. 1 Upper, extreme right), the FliIATPase regulator. This indicated that FliJ is not a generalsubunit chaperone.

FliJ Binds Chaperones for the ‘‘Minor’’ Subunits of the Filament Capand Hook Junction. Binding of FliJ to other flagellar componentswas screened by similar, glutathione Sepharose affinity copuri-fication. This revealed that GST-FliJ recognizes the minorsubunit-specific export chaperones FlgN and FliT (Fig. 1 Lower).In contrast, there was no recognition of a functional FliSchaperone [i.e., that bound its cognate subunit FliC subunit (2)],even when greater amounts of protein were tested or whenHis-tagged FliJ was used as bait in cell lysates, as in Fig. 1 Upper.

The FliJ-chaperone complexes were assembled in vitro andanalyzed by gel-filtration chromatography. The three chaperonesand FliJ each migrated alone as single species (Fig. 2 Ai–Aiii), thebehavior of the 20-kDa (His6)FliJ protein being consistent withthe aberrant migration of the putatively elongated monomerpreviously established by multiangle light scattering and analyt-ical ultracentrifugation (13). Following incubation with FliJ, theFlgN chaperone shifted in its elution from 30 kDa to �42 kDa(Fig. 2 Ai); similarly, FliT shifted from �25 kDa to �60 kDa (Fig.2Aii). By contrast FliS did not shift, remaining at �16 kDa (Fig.

Author contributions: L.D.B.E., G.M.F., and C.H. designed research; L.D.B.E., G.P.S., S.A., andG.M.F. performed research; L.D.B.E. and C.H. analyzed data; and L.D.B.E., G.P.S., G.M.F., andC.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission. O.S. is a guest editor invited by the Editorial Board.

Abbreviation: ITC, isothermal titration calorimetry.

*To whom correspondence should be addressed at: University of Cambridge, Downing Site,Tennis Court Road, Cambridge, Cambridgeshire CB2 1QP, United Kingdom. E-mail:ch@mole.bio.cam.ac.uk.

© 2006 by The National Academy of Sciences of the USA

17474–17479 � PNAS � November 14, 2006 � vol. 103 � no. 46 www.pnas.org�cgi�doi�10.1073�pnas.0605197103

2Aiii). The results are consistent with assembly of FlgN-FliJ andFliT-FliJ complexes but not of FliS-FliJ, and this was confirmedby immunoblotting of the fractions (not shown). Although theFliJ-FlgN elution would be compatible with a 1:1 stoichiometry,this method is not appropriate for accurate stoichiometry as-sessment, especially as FliJ migrates aberrantly. Incubation ofFliT with FlgN and FliJ together gave a single major elution peakat �62 kDa (Fig. 2 Aiv), which immunoblotting showed containsFliJ and both chaperones (Fig. 2 Aiv Lower). The location of thepeak is not compatible with a mixture of binary complexes,indicating that a ternary complex can be formed.

Identification of the ternary FliJ-FlgN-FliT complex sug-gested that the chaperones can occupy the 147 residue FliJsimultaneously, presumably at distinct sites. To assess this,(His10)FliJ variants containing scanning 10-residue deletions(13) were tested for binding to each of the FlgN and FliTchaperones in the Ni2�-affinity copurification assay. Theseassays revealed (Fig. 2B) that binding of FlgN, but not FliT,was abolished by FliJ deletions between residues 21 and 50,within the predicted N-terminal coiled-coil (13). Conversely,binding of FliT, but not FlgN, was disrupted by deletions in thecentral region between residues 61 and 100. FliJ variantscontaining any ‘‘nonbinding’’ deletion did not complement afliJ null mutant (13).

The results show that FlgN and FliT bind discrete nonover-lapping sites on FliJ and that these binding events are essentialfor export. This supports the size exclusion data, indicatinghow a ternary complex could be formed in the cell. Suchsimultaneous recruitment of both chaperones for all three

minor subunits in a ternary complex may be significant in FliJfunction. Our confirmation that FliJ interacts with FliH (Fig.1) (13, 14) indicates that FliJ may function at the membraneATPase complex. FliJ might recognize transient chaperone-subunit complexes en route to FliI docking, during docking, oreven after docking but before subunit release. Therefore, weexamined FliJ recognition of subunit-laden chaperones andthe location of FliJ.

FliJ Recognizes Only Free Chaperones and Is Associated with theMembrane. Chaperone-subunit complexes are transient in vivo(5), so FlgN-FlgK, FliT-FliD, and FliS-FliC complexes were

Fig. 1. Recognition of flagellar subunits and chaperones by FliJ. Affinitychromatography of (Upper) cell lysates of E. coli C41 expressing (His6) FliJ (20kDa) incubated with lysates of the same strain expressing one of the hook-classsubunits FlgD (24 kDa), FlgE (42 kDa, apparent molecular mass of 36 kDa), orFliK (42 kDa); the filament-class subunits FlgM (10 kDa), FlgK (59 kDa), FlgL (34kDa), FliD (50 kDa), or FliC (51 kDa); or the ATPase regulator FliH (26 kDa) and(Lower) (GST) FliJ protein (41 kDa) incubated with chaperones FlgN (18 kDa),FliT (16 kDa), or FliS (16 kDa), all affinity purified (3, 7). After incubation withresin (Ni2� or GST as appropriate), unbound (free, F) and bound (B) proteinswere separated by SDS�15% PAGE and stained with Coomassie blue.

Fig. 2. Assembly of FliJ-chaperone complexes. (A) Gel-filtration chromatog-raphy after in vitro incubation of purified FliJ with FlgN (i), FliT (ii), FliS (iii), andboth FlgN and FliT (iv) chaperones. Proteins were monitored spectrometeri-cally at A280. Molecular-mass protein standards are indicated above. Elutionfractions of the tripartite FliJ-FlgN-FliT complex were subjected to SDS�15%PAGE and immunoblotting with anti-FlgN (N) and anti-FliT sera (T), andHis-antibody (J). (B) Recognition of FlgN and FliT chaperones by (His10) FliJvariants containing 10-aa internal deletions. Affinity chromatography andSDS�PAGE were performed as in Fig. 1. Aligned below is a schematic of the147-residue FliJ protein indicating the FlgN and FliT binding sites.

Evans et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17475

MIC

ROBI

OLO

GY

preassembled in vitro following purification of proteins by af-finity chromatography (3, 7). FliJ binding was assessed by GSTaffinity chromatography and analyzed by SDS�PAGE (Fig. 3A).None of the assembled chaperone-subunit complexes copurifiedwith GST-FliJ.

We assessed whether chaperone-subunit complexes were rec-ognized in vivo by FliJ, assaying FliJ pull-down of FlgN-FlgKfrom S. typhimurium �fliJ�flgM in which class III f lagellar genesare expressed at all stages of flagellum assembly (15). Cellsexpressing (His10)FliJ were disrupted using a French pressurecell to release membrane-associated flagellar components to thesoluble fraction (7). From this release fraction (His10), FliJ wasaffinity purified, and coeluted proteins were assayed by immu-noblotting. The subunit FlgK was not detected (Fig. 3B), butFlgN chaperone was consistently copurified with FliJ. The datashowed that FliJ is unable to bind subunits whether chaperonedor not; they appear to preclude FliJ from being a cytosolic‘‘suprachaperone’’ or, indeed, acting to release chaperone-docked subunits at the export apparatus. FliJ has been surmisedto be a cytosolic chaperone (12), but the data (Fig. 1) (13)indicate FliJ interacts with FliH, which is complexed with FliI atthe inner membrane (7). Indeed, FliJ has also been reported to

interact with two other membrane-associated flagellar proteins:the export protein FlhA (13) and the C-ring protein FliM (16).To establish the location of FliJ, fractionation was carried out oncell lysates of a Salmonella �fliJ mutant producing complement-ing levels of (His10)FliJ. Immunoblotting with His-tag antiserarevealed FliJ was predominantly (�70%) in the membrane�insoluble fraction, with the lower proportion in the cytosol (Fig.4A). The same result was obtained with a Salmonella flhDC nullmutant producing (His10)FliJ in the absence of all other flagellarcomponents. This suggests that like the membrane-associatedFliH and FliI (7), FliJ has intrinsic membrane affinity.

Cell membrane fractions were separated by density gradientcentrifugation (17), and immunoblotting showed that (His10)FliJcolocalized with inner membrane fractions containing NADHoxidase activity and the FliH component of the ATPase complex(Fig. 4B). FliJ membrane affinity was also confirmed by an invitro liposome flotation assay (Fig. 4C), in which FliJ colocalizedwith E. coli phospholipids in a manner analogous to the inner-membrane-associated ATPase FliI and its regulator FliH, withwhich FliJ interacts (Fig. 1) (13, 14), and in contrast to thecytosolic chaperone FlgN (5). FliJ interaction with the ATPasecomplex is also consistent with a 2.4-fold increase observed in invitro FliI ATPase activity in the presence of FliJ (2.3–5.5�mol�min�1�mg�1 � �3%), whereas ATPase activity was unaf-fected by the FliJ-FlgN complex.

The accruing data had excluded a number of possibilities andsuggested that FliJ might recruit newly unloaded chaperones atthe ATPase complex (i.e., after subunits had been released intothe translocation phase). Such an activity could increase the localchaperone concentration at the export machinery, effecting acycle to enhance chaperone binding of new subunits and, thus,docking. To assess whether FliJ could direct such a chaperonetransfer and cycle, an in vitro chaperone capture assay wasdeveloped.

Subunit Capture of Cognate Chaperone from FliJ-Chaperone Com-plexes. Preformed (His6)FliJ-FlgN complex was bound to nickelbeads and challenged with cognate (FlgK) subunit. Immuno-blotting of FlgN that was released into the eluate or retained inthe FliJ-FlgN complex showed that FlgK displaced FlgN fromFliJ in a concentration-dependent manner (Fig. 5A), with �0.05�M FlgK sufficient to reduce by 50% the amount of FlgNretained by FliJ and 10 �M FlgK dislodging �90% of thechaperone. In parallel assays the (His6)FliJ-FlgN complex waschallenged by the same concentrations of the noncognate sub-unit FliC. By contrast, FliC was unable to displace the FlgNchaperone from FliJ (Fig. 5A).

To show that the putative FlgK subunit-FlgN chaperonecomplexes were formed after chaperone capture from FliJ,gel-filtration chromatography was performed on elution frac-tions from an assay in which 10 �M FlgK was applied. The results(Fig. 5B) show that in the elution peak the FlgN chaperone andFlgK subunit coeluted with an apparent molecular mass of 70

Fig. 3. Recognition of chaperone-subunit complexes by FliJ. (A) Affinitychromatography of (GST) FliJ (41 kDa) incubated with in vitro preformedchaperone-subunit complexes FlgN-FlgK (FlgNK), FliT-FliD (FliTD), and FliS-FliC(FliSC), assayed by SDS�PAGE and stained as in Fig. 1. (B) Affinity chromatog-raphy of whole cell (W) soluble lysates from S. typhimurium �fliJ�flgMexpressing (His10) FliJ, obtained by mechanical lysis. After incubation with Ni2�

agarose, the free (F) and bound (B) fractions were immunoblotted withanti-His (for FliJ), anti-FlgN, and anti-FlgK sera.

Fig. 4. Membrane localization of FliJ. (A) Cytosolic (cyt) and membrane�insoluble (mem) fractions of S. typhimurium �fliJ and �flhDC mutants expressing (His10)FliJ at �fliJ-complementing levels. (B) Membrane fractions of S. typhimurium fliJ [(His10) FliJ] separated on a 0.8–2.0 M sucrose gradient (75,000 � g, 16 h).Calibrated by Coomassie blue staining of outer membrane proteins (OMP) and inner membrane (IM) NADH oxidase activity (assay not shown), andimmunoblotted for FliJ with anti-His antibody and inner-membrane-associated FliH. (C) Lipid association of FliJ. Purified FliJ, peripheral membrane proteins FliIand FliH, and cytosolic FlgN were incubated separately with E. coli total phospholipid liposomes and placed at the bottom of a three-step sucrose density gradientand centrifuged (75,000 � g, 16 h). Proteins were precipitated from the top (T), middle (M), and bottom (B) fractions of the gradient before immunoblotting.

17476 � www.pnas.org�cgi�doi�10.1073�pnas.0605197103 Evans et al.

kDa, with residual FlgK monomer (49 kDa) and FlgN dimer(�25 kDa) also observed in the elution profile. The data showthat FliJ-bound chaperones are transferred to their cognatefilament subunits without requiring other proteins. As FliJ bindsonly empty chaperones and not subunits, this suggests that itsdisplacement involves either transient binding of chaperone byboth FliJ and subunit at different chaperone sites, effecting anallosteric change, or competition for the same binding site on thechaperone, with the subunit having greater affinity. We exam-ined these possibilities by establishing the sites and strengths ofthe two chaperone interactions.

Chaperone Transfer Is Affected by Subunit Displacement of FliJ froma Common Binding Site. To identify the region(s) of FlgN requiredfor binding of FliJ and the two subunits FlgK and FlgL, variantsof FlgN containing 10-residue scanning deletions in the C-terminal subunit-binding region were constructed and tested inaffinity copurification assays (Fig. 6A). All the variants wereexpressed at levels comparable to the full length and wereidentically soluble (input, Fig. 6A). In contrast to the full-lengthFlgN and most deleted variants, which bound all three proteins,FlgN variants lacking residues 81–90 and 91–100 abolishedbinding of both subunits and of FliJ. This is consistent with bothproteins binding to a common site on the chaperone. We soughtto establish the relative affinity of the competing FliJ and FlgKproteins for the chaperone FlgN using isothermal titrationcalorimetry (ITC). The heat output measured over time and thederived integrated measure of heat transfer with respect toconcentration depicted in both cases an exothermic interactionthat allowed derivation of affinity values using a single-binding-site model. Stepwise titration of FlgN into FlgK (Fig. 6B)revealed that chaperone bound to subunit with a KD of 0.03 �M.Similar titration of FlgN into FliJ showed binding to be morethan 700-fold weaker, with a KD of 22 �M. This indicates thatchaperone transfer to subunit is achieved by direct competitionand displacement at a common binding site.

DiscussionOur screen for FliJ intermediate export complexes revealed thatFliJ recruits unladen FlgN and FliT, which chaperone the minorfilament-class subunits FlgK, FlgL, and FliD, but not with FliS,which chaperones the major component FliC (flagellin). FliJformed binary and ternary complexes with the chaperones viadistinct, binding sites that do not overlap the region determininginteraction with FliH, the FliI ATPase regulator. Deletionswithin either chaperone binding site disabled flagellar subunitexport (13). FliJ did not bind chaperone-subunit complexes,consistent with chaperoned subunit docking occurring in theabsence of FliJ (5). This indicates that FliJ is not involved intransition of chaperoned subunits from the cytosol to themembrane. Like the ATPase complex components FliI and FliH(7), FliJ localizes predominantly to the inner membrane and hasintrinsic affinity for phospholipids; it interacts with FliH (13, 14)and also the membrane export component FlhA (13) and C-ringprotein FliM (16). FliJ increased FliI ATPase activity in vitro, but

Fig. 5. Capture of FliJ-bound chaperone by cognate subunit. (A) FliJ-boundand subunit-eluted FlgN was assayed by immunoblotting following incuba-tion of Ni2� agarose-bound in vitro preformed (His6) FliJ-FlgN chaperonecomplex with subunit, either cognate FlgK (Upper) or noncognate FliC(Lower). (B) Gel filtration (S200) of the putative chaperone-subunit complexcaptured in A by 10 �M FlgK subunit, monitored spectrometerically at A280.Elution fractions were immunoblotted with anti-FlgK (subunit) and anti-FlgN(chaperone) sera.

Fig. 6. Competitive binding of a common site on the chaperone by cognatesubunit and FliJ. (A) Binding of FliJ and cognate subunits to the chaperoneFlgN. Ni2� affinity chromatography (as in Fig. 1) and Coomassie blue stainingindicate recognition of internally deleted FlgN variants (input, all expressedand soluble comparable to the full-length FlgN) by His-tagged FliJ, FlgK, orFlgL. Aligned below is a depiction of the coincident binding sites on the140-residue FlgN. (B) Isothermal calorimetry (ITC) of FlgN chaperone bindingto FliJ (Left) and to subunit FlgK (Right). Heat uptake upon stepwise injection(1 � 1 �l; then 18 � 3.5 �l) of FlgN (360 �M) into the calorimetric cell (1.4 ml)containing FliJ (73 �M) (Left) and injection of (1 � 1 �l; then 29 � 4 �l) of FlgN(260 �M) into the calorimetric cell containing FlgK (23 �M) (Right). Heat pulsesin the absence of FlgN were negligible. (B Lower) Integrated heat pulses,which were normalized per mole of injectant, showing differential bindingcurves.

Evans et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17477

MIC

ROBI

OLO

GY

it does not copurify with FliI in affinity chromatography (notshown), suggesting that there might be a FliJ-FliI interaction thatis transient and�or stabilized by other flagellar components.These data indicated that FliJ had a novel activity, binding emptychaperones at the export ATPase complex. By establishing an invitro capture assay we were able to show that cognate (but notnoncognate) subunits can sequester chaperones from FliJ com-plexes in a concentration-dependent manner. FliJ was unable tosequester chaperones from cognate subunit. FliJ and the sub-units FlgK and FlgL bind the same 20-residue region of the FlgNchaperone, pointing to direct competition at a common bindingsite. This was confirmed by ITC, which revealed that FlgNchaperone has a �700-fold higher affinity for its cognate subunitFlgK than for FliJ. These data appear to explain how emptychaperones are recruited and held by FliJ before transfer to theircognate subunits.

Our findings offer a new view of the sequence of eventsunderlying export. We envisage that chaperone-piloted subunitsdock transiently at the membrane ATPase complex, whichcatalyzes subunit unfolding and translocation, and effects chap-erone release. FliJ then acts as a chaperone escort protein,clearing the ATPase of unloaded chaperones, which it transfersto new subunits. This view is compatible with our observedstimulation of FliI ATPase activity by free FliJ but not FliJ-chaperone complex. This escort activity of FliJ at the exportATPase complex would provide a chaperone sink to enhanceexport by allowing cycling of chaperones to increase the fre-quency of productive subunit-chaperone complex formation anddocking. Our data reveal an entirely novel function for FliJ in thetype III export of flagellar subunits: that of chaperone escort.

Whether such a chaperone escort mechanism occurs in therelated type III systems that secrete virulence effectors is notobvious, but one can identify genes (e.g., Salmonella invI�spaM,Yersinia yscO, Shigella spa13, and E. coli orf15) immediatelydownstream of the export ATPase gene that encode proteins ofsimilar size (18–20 kDa) and charge to FliJ.

The current view of the late stage of flagellar assembly doesnot encompass any notion of filament-class subunit exporthierarchy. However, the flagellum stoichiometry suggests thatminor components of the hook-filament junction (11 monomersof FlgK and FlgL per filament) and the filament cap (5 mono-mers of FliD) may have to compete for export with an excess offlagellin monomers. This view is strengthened by measurementsthat indicate comparable ratios of unincorporated subunits inthe extra cellular medium (18) and cytosol (our unpublisheddata). This apparent problem would be compounded by the needfor all three minor substructures to assemble before flagellin isincorporated (19), and suggests that there might be a mechanismto favor export of the minor subunits, especially if successiverounds of selection are required once each substructure beginsto assemble. Selectivity could best act at individual nascentassemblies during or after docking. Our data could indicate abasis for such a selectivity process, in which FliJ recruits andescorts chaperones for the minor subunits as part of a cycle topromote the early formation of the adjacent hook-filamentjunction and cap structures.

MethodsBacterial Strains and Plasmids. Bacteria were cultured at 37°C tolate exponential phase (A6001.0), unless stated, in Luria-Bertani(LB) broth containing, where appropriate, ampicillin, chloram-phenicol, or kanamycin (each at 20 �g�ml�1). Wild-type S.typhimurium SJW1103 is motile (20), and derived mutants carrylesions in the cheW-flhD locus [SJW1368 (21)]. Mutants�fliJ::Km

R, �flgM::KmR, and �fliJ::FRTflgM::Km

R, in which geneswere replaced by a kanamycin resistance cassette, were con-structed using P22 transduction and�or the � Red recombinasesystem (22). Recombinant proteins were expressed in E. coli C41

(23) from IPTG inducible plasmids pET15b (24), pACT7 (25),pGEX-4T-3 (26), or pTrc99-FF4 (27). To construct recombinantplasmids encoding, individual f lagellar genes S. typhimuriumflgD, flgE, flgK, flgL, flgM, flgN, fliC, fliD, fliH, fliJ, fliK, fliS, andfliT were each amplified from chromosomal DNA by PCR usingPfu turbo DNA polymerase. PCR products were inserted NdeI�BamHI into pACT7 or pET15b. For GST fusion constructs,genes were amplified by PCR, and products inserted BamHI�XhoI into pGEX-4T-3. FlgN internal 10-residue deletions werecreated by overlap-extension PCR and products inserted NdeI�BamHI into pACT7. Inserts were verified by DNA sequencing(Department of Genetics, University of Cambridge).

Purification of Proteins. (His6)-tagged proteins FlgN, FliT, FliS,and FliJ were purified as described (2, 3). Chaperone-subunitcomplexes were prepared from cells coexpressing (His6)-tagged chaperone and untagged subunit. Cells resuspended inbuffer A (50 mM NaH2PO4, pH 7.4�150 mM NaCl�1 mMDTT�5 mM imidazole) were lysed by French pressure cell(Aminco). Membranes, unlysed cells, and insoluble proteinswere removed by centrifugation (40,000 � g, 1 h), and thecleared cell lysate was passed over nickel nitrilotriacetic acid(Ni2�) agarose (Qiagen). After washing with 50 column vol-umes (buffer A with 30 mM imidazole), complexes were eluted(buffer A with 700 mM imidazole), dialyzed against buffer A,and further purified by Superdex 200 (GE Healthcare) gelfiltration. FliI was purified as described (8).

Affinity Chromatography Copurification Assays. Copurification(pull-down) of protein complexes was achieved with either Ni2�

agarose or glutathione Sepharose 4B (3, 5). In vitro mixedpurified proteins or cleared cell lysates were incubated for 1 hwith affinity resin. After extensive washing [buffer A (�10–60mM imidazole)], proteins were eluted in buffer A containingeither 700 mM imidazole or 20 mM glutathione (no detergentwas used in these assays). For in vivo studies, soluble lysates ofS. typhimurium strains expressing (His10)FliJ at a complement-ing level were prepared as above, incubated for 1 h with Ni2�,washed three times with buffer A (60 mM imidazole), andproteins eluted in SDS loading buffer. Using untagged FliJ wasprecluded by nonspecific binding to resin.

Analytical Gel-Filtration Chromatography. Gel-filtration chroma-tography used Superdex 200 HR 10�30 (GE Healthcare). Pro-tein samples (0.1–5 mg�ml�1, 1% bed volume) were resolved ata flow rate of 0.5 ml�min�1; 0.3 ml elution fractions wereprecipitated [10% (wt�vol) TCA] and resuspended in SDSloading buffer.

Salmonella Cell Fractionation and Membrane Separation. Cultureswere separated into membrane and cytosolic fractions by chem-ical lysis (5, 7). Membranes were analyzed on a 16-ml stepwisesucrose gradient (1–2 M) centrifuged for 16 h at 75,000 � g (17).

Liposome Flotation Assay. Purified proteins (1–2 �g) were mixedwith 40 �l of 10 mg�ml�1 E. coli total phospholipids (Avanti PolarLipids) in TBS (20 mM Tris, pH 7.4�150 mM NaCl) (7) andincubated at room temperature for 15 min. Sucrose was addedto 50% (wt�vol in 1 ml of TBS), and samples were overlaid with3.5 ml of 40% sucrose (wt�vol in TBS) and 0.5 ml of TBS. Aftercentrifugation at 75,000 � g for 16 h at 16°C, 10 fractions (0.5 ml)were collected and precipitated [10% (wt�vol) TCA]. Fractions1–4, 5–7, and 8–10 were pooled to give top (T), middle (M), andbottom (B) fractions, respectively.

Steady-State ATPase Assay. FliI ATP hydrolysis activity was mea-sured at A340 by enzyme-coupled ATP�NADH oxidation (28) at37°C in reaction buffer (50 mM triethanolamine, pH 8.0�10 mM

17478 � www.pnas.org�cgi�doi�10.1073�pnas.0605197103 Evans et al.

magnesium acetate�1 mM DTT) in the presence of 1 mMphosphoenolpyruvate (PEP), 0.15 mM NADH, pyruvate kinase(5 units), lactate dehydrogenase enzymes (3.5 units) (rabbitmuscle; Sigma), 0.1 mg�ml E. coli total phospholipids, ATP(3–0.01 mM), and FliJ (0.1–3 �M). Reactions were initiated withthe addition of FliI (50 �g).

ITC. ITC was performed at 25°C using a VP-ITC system andOrigin software (Microcal Inc). Proteins were dialyzed into 50mM NaH2PO4 (pH 7.4), 150 mM NaCl, and 1 mM DTT. The

heat evolved following each injection was obtained from theintegral of the calorimetric signal. The heat due to the bindingreaction is the difference between the heat of reaction and theheat of dilution.

We thank the Macnab laboratory (Yale University, New Haven, CT) forproviding strains and plasmids and R. Hayward for critical reading of themanuscript. This work was supported by a Wellcome Trust Programgrant (to C.H.) and a Commonwealth Scholarship Commission Student-ship (to S.A.).

1. Macnab RM (2003) Annu Rev Microbiol 57:77–100.2. Auvray F, Thomas J, Fraser GM, Hughes C (2001) J Mol Biol 308:221–229.3. Bennett JC, Thomas J, Fraser GM, Hughes C (2001) Mol Microbiol

39:781–791.4. Fraser GM, Bennett JC, Hughes C (1999) Mol Microbiol 32:569–580.5. Thomas J, Stafford GP, Hughes C (2004) Proc Natl Acad Sci USA 101:3945–

3950.6. Gauthier A, Finlay BB (2003) J Bacteriol 185:6747–6755.7. Auvray F, Ozin AJ, Claret L, Hughes C (2002) J Mol Biol 318:941–950.8. Claret L, Calder SR, Higgins M, Hughes C (2003) Mol Microbiol 48:1349–1355.9. Samatey FA, Imada K, Nagashima S, Vonderviszt F, Kumasaka T, Yamamoto

M, Namba K (2001) Nature 410:331–337.10. Akeda Y, Galan JE (2005) Nature 437:911–915.11. Minamino T, Macnab RM (1999) J Bacteriol 181:1388–1394.12. Minamino T, Chu R, Yamaguchi S, Macnab RM (2000) J Bacteriol 182:4207–

4215.13. Fraser GM, Gonzalez-Pedrajo B, Tame JR, Macnab RM (2003) J Bacteriol

185:5546–5554.14. Gonzalez-Pedrajo B, Fraser GM, Minamino T, Macnab RM (2002) Mol

Microbiol 45:967–982.15. Gillen KL, Hughes KT (1991) J Bacteriol 173:2301–2310.

16. Gonzalez-Pedrajo B, Minamino T, Kihara M, Namba K (2006) Mol Microbiol60:984–998.

17. Osborn MJ, Munson R (1974) Methods Enzymol 31:642–653.18. Komoriya K, Shibano N, Higano T, Azuma N, Yamaguchi S, Aizawa SI (1999)

Mol Microbiol 34:767–779.19. Homma M, Kutsukake K, Iino T, Yamaguchi S (1984) J Bacteriol 157:100–108.20. Yamaguchi S, Fujita H, Taira T, Kutsukake K, Homma M, Iino T (1984) J Gen

Microbiol 130:3339–3342.21. Ohnishi K, Ohto Y, Aizawa S, Macnab RM, Iino T (1994) J Bacteriol

176:2272–2281.22. Datsenko KA, Wanner BL (2000) Proc Natl Acad Sci USA 97:6640–6645.23. Miroux B, Walker JE (1996) J Mol Biol 260:289–298.24. Studier FW, Moffatt BA (1986) J Mol Biol 189:113–130.25. Thanabalu T, Koronakis E, Hughes C, Koronakis V (1998) EMBO J 17:6487–

6496.26. Kaelin WG, Jr, Krek W, Sellers WR, DeCaprio JA, Ajchenbaum F, Fuchs CS,

Chittenden T, Li Y, Farnham PJ, Blanar MA, et al. (1992) Cell 70:351–364.27. Ohnishi K, Fan F, Schoenhals GJ, Kihara M, Macnab RM (1997) J Bacteriol

179:6092–6099.28. Trentham DR, Bardsley RG, Eccleston JF, Weeds AG (1972) Biochem J

126:635–644.

Evans et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17479

MIC

ROBI

OLO

GY