doi:10.1016/j.jmb.2007.09.080 J. Mol. Biol. (2007) 374, 877–882

Available online at www.sciencedirect.com

COMMUNICATION

Sorting of Early and Late Flagellar Subunits AfterDocking at the Membrane ATPase of the Type IIIExport Pathway

Graham P. Stafford†, Lewis D. B. Evans†, Rita Krumscheid,Paraminder Dhillon, Gillian M. Fraser and Colin Hughes⁎

Cambridge UniversityDepartment of Pathology,Tennis Court Road,Cambridge CB2 1QP, UK

Received 14 June 2007;received in revised form20 September 2007;accepted 27 September 2007Available online3 October 2007

*Corresponding author. E-mail addrch@mole.bio.cam.ac.uk.† G.P.S. and L.D.B.E. contributedPresent address: G. P. Stafford, De

Pathology, School of Clinical DentisSheffield, Claremont Crescent, Sheff

0022-2836/$ - see front matter © 2007 E

The bacterial flagellum assembles in a strict order, with structural subunitsdelivered to the growing flagellum by a type III export pathway. Early rod-and-hook subunits are exported before completion of the hook, at whichpoint a subunit-specificity switch allows export of late filament subunits.This implies that in bacteria with multiple flagella at different stages ofassembly, each export pathway can discriminate and sort unchaperonedearly and chaperoned late subunits. To establish whether subunit sorting isdistinct from subunit transition from the cytosol to the membrane, inparticular docking at the membrane-associated FliI ATPase, the pathwaywas manipulated in vivo. When ATP hydrolysis by the FliI ATPase wasdisabled and when the pathway was locked into an early export state, bothunchaperoned early and chaperoned late subunits stalled and accumulatedat the inner membrane. Furthermore, a chaperone that attenuates latesubunit export by stalling when docked at the wild-type ATPase also stalledat the ATPase in an early-locked pathway and inhibited export of earlysubunits in both native and early-locked pathways. These data indicate thatthe pathways for early and late subunits converge at the FliI ATPase,independent of ATP hydrolysis, before a distinct, separable sorting step. Toascertain the likely signals for sorting, the export of recombinant subunitswas assayed. Late filament subunits unable to bind their chaperones werestill sorted accurately, but chaperoned late subunits were directed throughan early-locked pathway when fused to early subunit N-terminal exportsignal regions. Furthermore, while an early subunit signal directed export ofa heterologous type III export substrate through both native and early-locked pathways, a late subunit signal only directed export via nativepathways. These data suggest that subunits are distinguished not by latechaperones but by N-terminal export signals of the subunits themselves.

© 2007 Elsevier Ltd. All rights reserved.

Keywords: flagella assembly; type III export; sorting mechanism; exportchaperone

Edited by I. B. Holland

Bacterial motility is commonly conferred by cellsurface flagella, comprising a long helical filament

ess:

equally to this work.partment of Oraltry, University ofield S10 2TA, UK.

lsevier Ltd. All rights reserve

that is connected by a flexible hook to a central rodin the cell envelope basal body that also housesthe flagellar motor.1–3 Flagella substructures areassembled in strict sequence, with formation of thebasal body and rod structures preceding polymer-isation of the hook and, finally, the filamentsubunits.1–3 The order of assembly is achieved bysequential expression of the gene hierarchy4,5 andby a subunit-specificity switch in the flagellar typeIII export pathway. This ensures that, prior to hook

d.

Fig. 1. Membrane accumulation of early and latesubunits in the pathway attenuated by enzymaticallyimpaired FliI ATPase. (a) FliC export, assayed byimmunoblotting, filtered supernatants from midexponen-tial Luria broth (LB) cultures (A600 = 1.0) of ΔfliIflgKflgMcells (made by P22 transduction combined with themethod of Datsenko and Wanner27) expressing in transeither wild-type FliI (FliIWT) or variant FliIE211A (FliIEA)from pBAD33 (0.1% arabinose). A ΔfliIflgKflgM straincontaining empty pBAD33 was shown to be nonmotileand attenuated in the export of early FliK subunit and lateFliC subunit (data not shown). (b) Salmonella fliIflgKflgMcultures expressing wild-type FliIWT or variant FliIEA

separated into membrane (m) and cytoplasmic (c)fractions.11,16 Immunoblotted for FliI ATPase, FlgNchaperone and subunits. (c) Separation of the membranefractions into outer membrane (OMP; Coomassie stained)and inner membrane (NADH oxidase marker) by sucrosegradient ultracentrifugation (0.8–2.0 M11,16 top andbottom of the gradient indicated). Proteins immuno-blotted using antisera described above.

878 Sorting of Early and Late Flagellar Subunits

completion, only ‘early’ rod-and-hook subunits areexported,6,7 while those forming the later distal sub-structures of the filament, filament cap and hook–filament junction are not. The outline of the subunit-specificity switching mechanism is evident. Whenthe hook reaches its mature length, a signal istransmitted by the hook length control protein FliKto the integral membrane export component FlhB,triggering a switch in subunit specificity to allowexport of late subunits.6–9 Nevertheless, peritri-chously flagellated bacteria such as Escherichia coliand Salmonella have multiple flagella at differentstages of assembly,10 so an individual export path-way potentially encounters both unchaperonedearly and chaperoned late subunits from the cytosol.This implies that early and late subunits are dis-criminated and sorted by the pathway.We have previously shown that late filament

subunits are piloted by their chaperones to dockat the membrane-associated FliI ATPase.11 Herewe manipulate the export pathway to determinewhether subunit docking and sorting are sepa-rable and sequential events. We also assess the rela-tive influence of subunit export signals and boundexport chaperones in discriminating early and latesubunits.

Stalling of early and late subunits at themembrane in an early-locked pathwayattenuated in ATP hydrolysis

To examine the relationship between the proposedsorting step and subunit transition from the cytosolto the inner membrane, we aimed to generate stalledexport intermediates of both early and late subunits.Our previous work had exploited export-defectivechaperones to stall late cognate (hook–filamentjunction) subunits, which they piloted to and dockedat the membrane-associated FliI ATPase.11 Tosimilarly interrupt the movement of unchaperonedearly subunits, we attenuated FliI ATP hydrolysis,which, as in other export systems,12,13 is envisagedto drive unfolding and export of substrates engagedat the membrane machinery,14 in this case prior toassembly into the growing flagellum. After creatingsingle-amino-acid substitutions in the active siteregion, one variant was chosen for full study, variantFliIE211A, which is mutated immediately adjacent tothe Walker A motif. ATP was still bound by FliIE211A[Km = 0.2 mM, compared to wild type (1 mM);triplicate assays ±15%] but was poorly hydrolysed[Vmax = 0.22 μmol min−1mg−1 compared to wildtype (2.30)] in a coupled assay in the presence ofphospholipids.15

Export supported by FliIE211A was assayed in afliIflgKflgM triple mutant (by our previously pub-lished method11). The resulting pathway is notsubject to negative feedback (via the FlgM anti-sigma factor) arising from the disabling of the exportapparatus (fliI ATPase), and late and early subunitsare thus constitutively synthesized. When wild-typefliI is expressed in trans, exported late subunits suchas filament subunit FliC accumulate in the culture

supernatant as a result of the flgK hook–junctionlesion that precludes filament polymerisation(Fig. 1a). Export of FliC was severely attenuatedby substitution of FliI by FliIE211A. Like the wild-typeATPase, FliIE211A assembled into hexamers in vitroin the presence of phospholipids and the short-arm crosslinker disuccinimidylglutamate (Supple-mentary Data), and cell fractionation and su-crose gradient ultracentrifugation11,16 showed that,in vivo, it localised normally to the inner membrane(Fig. 1b and c).

Fig. 2. Membrane accumulation of early and late sub-units in an early-locked pathway attenuated by stallingdocked chaperone FlgNrel. (a) Localisation of FlgNrel

(expressed in trans by 0.01% arabinose) in the whole cell(wc), membrane (m) and cytoplasm (c)11,15 of the ΔflgNandΔfliI pathways, and in the isogenic (expE) early-lockedpathways ΔflgEN and ΔflgEfliI. (b) Affinity copurificationof stalled docking complexes by His–FlgNrel bait (+)11

from native ΔflgN and early-locked (expE; ΔflgEN) path-ways [(−) vector-only controls]. Cell extracts wereincubated with Ni–NTA resin [20 mM tris(hydroxy-methyl)aminomethane–HCl pH 8.0, 300 mM NaCl and5 mM imidazole] before washing (10 mM imidazole) andelution in SDS sample buffer. FlgN chaperone, FlgKcognate subunit and ATPase complex components FliIand FliH were detected by immunoblotting.

879Sorting of Early and Late Flagellar Subunits

The fliIflgKflgM pathway containing the nonhy-drolysing FliIE211A is locked into an early exportstate.6,7,11 In vivo localisation of nonexported sub-units in this pathway revealed (Fig. 1b and c) thatthe unchaperoned early subunit FliK17 accumulatedas a membrane-associated intermediate in a FliI-dependent manner. This indicates that, like chaper-oned late subunits, unchaperoned early subunitscan be stalled at the membrane, putatively dockedat the FliI ATPase. If late subunits are sorted beforethey dock at FliI, then late subunit–chaperonecomplexes should not accumulate at the membranein the FliIE211A early-locked fliIflgKflgM pathway,but they should accumulate if sorting occurs afterlate subunit docking. The in vivo fractionation andsucrose gradients of fliIflgKflgM cells expressingFliIE211A (Fig. 1b and c) revealed that the late sub-units FliC and FlgL and the FlgN chaperone18,19

accumulate, like early FliK, at the inner membrane.The data indicate that FliI enzymatic activity is not

required for in vivo docking of late subunits at themembrane ATPase (compatible with in vitro inter-action of virulence chaperones with a catalyticallyinactive type III export ATPase14), and they indicatethat this is also true for unchaperoned earlysubunits. Furthermore, they argue that sorting isseparable from docking at FliI, occurring most likelyafterwards, and that progression to sorting requiresATP hydrolysis by FliI.

Early and late subunits converge at the ATPaseprior to sorting

We have described a late FlgN chaperone variant(now called FlgNrel, as it putatively fails to releasefrom the ATPase) that attenuates export of cognateand noncognate late subunits when expressed intrans in wild-type pathways, with chaperonedsubunits trapped after docking at the membraneFliI, accumulating chaperone–subunit–ATPaseintermediates.11 We used this dominant-negativechaperone variant to extend indications that chap-eroned late subunits engage the wild-type FliIATPase before sorting, asking whether FlgNrel-stalled membrane intermediates accumulate in anearly export-locked pathway and attenuate earlysubunit export.Cell fractionation was performed to establish

FlgNrel localisation and putative interaction withthe ATPase in the actively secreting early-lockedpathway lacking the hook protein FlgE and thewild-type FlgN chaperone (expE; ΔflgEΔflgN), andalso in native pathways of ΔflgN control (expE+L)that export early subunits and then late subunitsafter completion of the hook substructure. Theresults (Fig. 2a) show that FlgNrel localised to themembrane in a FliI-dependent manner in bothpathways. When His-FlgNrel was used as bait in invivo affinity chromatography (Fig. 2b), cognatesubunit FlgK, FliI and its regulator FliH were allcopurified, showing that ATPase–chaperone–subu-nit intermediate complexes were formed in early-locked pathways analogous to native pathways.

These findings support the earlier indication thatboth unchaperoned early and chaperoned latesubunits engage the export ATPase before they aresorted for export or exclusion, and furthermoresuggest that the export pathways for the two classesof subunit converge at the ATPase. To substantiatethe idea of convergence, we again used the dockedbut stalled late FlgNrel to assess whether it couldattenuate export of not only chaperoned latesubunits but also early subunits in native exportpathways of ΔflgN. The results (Fig. 3) confirm thatexport of cognate (FlgK) and noncognate (FliC)chaperoned late subunits is reduced by 5- to 10-foldand reveal a comparable attenuation in the export ofunchaperoned early subunits FlgD and FliK. Sig-nificantly, they show that FlgNrel caused a compar-able reduction of FlgD and FliK export in an earlyexport-locked (ΔflgEΔflgN; expE) pathway (Fig. 3).These FlgNrel experiments strengthen the indicationfrom those using FliIE211A, that is, that the pathwaysfor unchaperoned early and chaperoned late sub-units converge at the ATPase before progressing(dependent on ATP hydrolysis) to sorting.

Are subunits discriminated by their own signalsor by late chaperones?

An obvious difference between early and latesubunits is that only late subunits are bound byexport chaperones, acting as cytosolic bodyguardsand pilots for docking at the membrane exportATPase.11,20,21 Chaperones could act as flagellar

Fig. 3. Attenuation of early and late subunits exportby stalling FlgNrel. Export of subunits by native ΔfliD(expE+L) and early-locked ΔflgE (expE) pathways contain-ing FlgNrel [expressed using 0.01% arabinose; (−) vector-only controls], assayed following precipitation fromsupernatants (snt) of midexponential LB cultures (wc,whole culture) by SDS-PAGE and immunoblotting forearly (FliK and FlgD) and late (FliC and FlgK) subunits.

880 Sorting of Early and Late Flagellar Subunits

sorting signals, labelling subunits for rejectionduring the early stages of flagella assembly. This isespecially so as in a virulence type III secretionsystem, bound chaperones (e.g., Salmonella InvBchaperone of the SopE effector), are reported to formpart of the secretion signal, preventing promiscuousexport through the flagellar pathway.20,21 If thiswere true, late subunits from which C-terminalpolymerisation and chaperone-binding domains aredeleted might be exported as early subunits. To test

amino acids 1–100) fused to the signal-less SptP tyrosine phosp(expE; ΔflgE) and native (expE+L; ΔfliC) pathways. ProteinsUniversity of Cambridge). Genes encoding variant wild-typeoverlap extension PCR using Salmonella chromosomal DNA arestriction sites of the pBAD18 expression vector. Recombinanwere assayed as in Fig. 2b. Control experiments performedtransduction of ΔfliI allele into ΔflgE) showed that none of th

this possibility, we assessed the export of recombi-nant FliC and FlgK late subunit variants lackingtheir chaperone-binding domains (FlgKΔchap, andFliCΔchap)

6,18,22 and found (Fig. 4a) that, althoughthese variants were exported in native pathways thatexport both early and late subunits (expE+L), neithervariant was exported in an early-locked pathway(expE; i.e., the unchaperoned subunits were stillfaithfully sorted by the export pathway). We thenassessed the export of recombinant subunits inwhich the N-terminal export signal residues 1–100(FlgDsig) were fused to truncated late FliC (FliCΔsig)or FlgK (FlgKΔsig) lacking their N-terminal exportsignals23 but still able to bind their respectivechaperones. Like wild-type FlgD, these hybridsubunits were exported by both early export-locked(expE) and native export (expE+L) pathways (Fig.4b), substantiating the view that chaperones are not asorting signal to preclude late subunit export beforehook completion. The marginal (twofold) reductionin the export of early subunits (FlgD, FlgDsig–FliCΔsig, FlgDsig–FlgKΔsig) in native pathways (Fig.4b; expE+L), compared to the early export-lockedstrain (expE), possibly indicates that once an exportapparatus has switched specificity, it no longeraccepts early subunits for export (a view compatiblewith observations in Yersinia T3SS indicating thatonce pathways have switched specificity to lateeffectors, they do not export early substrates24).

Fig. 4. Influence of subunit do-mains on sorting. (a) Export ofchaperoned late subunits (FlgKand FliC) and their variants thatcannot bind chaperone (FlgKΔchapand FliCΔchap) in export pathwaysthat are either early locked (expE;ΔflgEKL27 ΔflgE, SJW1353 acquiredfrom Ohnishi et al.28) or native(expE+L; ΔflgKL or ΔfliC). Proteinsfrom whole cells (wc) and super-natants (snt) were immunoblottedwith FlgK or FliC antisera. (b)Export of FlgD and recombinantlate subunits FlgKΔsig and FliCΔsiglacking amino acids 1–100 fused toresidues 1–100 of early FlgD inexport pathways (described above)that are either early locked (expE) ornative (expE + L). Proteins fromwhole cells (wc) and supernatants(snt) were immunoblotted withFlgD, FlgK or FliC antisera. (c)Export of recombinant fusion pro-teins comprising putative early orlate subunit N-terminal signalregions (FlgDsig, FlgKsig and FliCsig;

hatase domain (SptPphos; residues 161–543) in early lockedwere immunoblotted with SptP antisera (V. Koronakis,and variant FliC, FlgK, FlgD and SptP were amplified bys template. PCR products were inserted into XbaI–HindIIIt genes were expressed (LB, 0.01% arabinose) and proteinsin isogenic ΔfliI and ΔflgEfliI strains (created by P22

e recombinant proteins was exported (data not shown).

881Sorting of Early and Late Flagellar Subunits

These assays also suggest that the subunit N-terminal 100 residues contain both export andsorting signals. To confirm this, we constructedrecombinant subunits comprising the N-terminal100 residues of the late subunit FliC (FliCsig) or FlgK(FlgKsig), or the early subunit FlgD (FlgDsig) fused tothe catalytic phosphatase domain of the SalmonellaSPI-1 SptP effector (residues 161–543; SptPphos)

14

and assayed export in early export-locked and wild-type equivalent pathways. As expected, FlgDsig–SptPphos fusion was exported in both early-locked(expE) and wild-type (expL+E) pathways (Fig. 4c). Incontrast, the FliCsig–SptPphos and FlgKsig–SptPphosfusion proteins were exported only in the wild-typepathway (expE+L) and not by the early export-locked pathway (expE) (Fig. 4c). This supportsour view that the N-terminal regions contain suf-ficient information to determine sorting as a latesubunit. While there is primary sequence similarityamong the N-terminal regions of rod subunits, thereseems to be little identity between these and otherearly subunits25,26 and no obvious identity distin-guishing the N-terminal regions of late proteins.This suggests that subunit sorting might rely on therecognition of structural features specific to eachsubunit class.

Acknowledgements

We thank Vasillis Koronakis for SptP antisera. Thiswork was supported by a Wellcome Trust Pro-gramme grant (C.H.) and a Biotechnology andBiological Sciences Research Council studentship(P.D.).

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2007.09.080

References

1. Macnab, R. M. (1996). Flagella and motility. In Escher-ichia coli and Salmonella typhimurium: Cellular andMolecular Biology (Neidhart, F. C., ed), pp. 123–145,American Society for Microbiology, Washington, DC.

2. Macnab, R. M. (2003). How bacteria assemble flagella.Annu. Rev. Microbiol. 57, 77–100.

3. Namba, K. & Vonderviszt, F. (1997). Moleculararchitecture of bacterial flagellum. Q. Rev. Biophys.30, 1–65.

4. Kutsukake, K., Ohya, Y. & Iino, T. (1990). Transcrip-tional analysis of the flagellar regulon of Salmonellatyphimurium. J. Bacteriol. 172, 741–747.

5. Soutourina, O. A. & Bertin, P. N. (2003). Regulationcascade of flagellar expression in Gram-negativebacteria. FEMS Microbiol. Rev. 27, 505–523.

6. Fraser, G. M., Hirano, T., Ferris, H. U., Devgan, L. L.,Kihara, M. & Macnab, R. M. (2003). Substratespecificity of type III flagellar protein export in

Salmonella is controlled by subdomain interactions inFlhB. Mol. Microbiol. 48, 1043–1057.

7. Hirano, T., Minamino, T., Namba, K. & Macnab, R. M.(2003). Substrate specificity classes and the recogni-tion signal for Salmonella type III flagellar export.J. Bacteriol. 185, 2485–2492.

8. Hirano, T., Yamaguchi, S., Oosawa, K. & Aizawa, S.(1994). Roles of FliK and FlhB in determination offlagellar hook length in Salmonella typhimurium.J. Bacteriol. 176, 5439–5449.

9. Hirano, T., Shibata, S., Ohnishi, K., Tani, T. & Aizawa,S. (2005). N-terminal signal region of FliK is dispen-sable for length control of the flagellar hook. Mol.Microbiol. 56, 346–360.

10. Aizawa, S. I. & Kubori, T. (1998). Bacterial flagellationand cell division. Genes Cells, 3, 625–634.

11. Thomas, J., Stafford, G. P. & Hughes, C. (2004).Docking of cytosolic chaperone–substrate complexesat the membrane ATPase during flagellar type IIIprotein export. Proc. Natl. Acad. Sci. U. S. A. 101,3945–3950.

12. Schiebel, E., Driessen, A. J., Hartl, F. U. & Wickner, W.(1991). Delta mu H+ and ATP function at differentsteps of the catalytic cycle of preprotein translocase.Cell, 64, 927–939.

13. Thanabalu, T., Koronakis, E., Hughes, C. & Koronakis,V. (1998). Substrate-induced assembly of a contiguouschannel for protein export from E. coli: reversiblebridging of an inner-membrane translocase to anouter membrane exit pore. EMBO J. 17, 6487–6496.

14. Akeda, Y. & Galan, J. E. (2005). Chaperone release andunfolding of substrates in type III secretion. Nature,437, 911–915.

15. Evans, L. D., Stafford, G. P., Ahmed, S., Fraser, G.M. &Hughes, C. (2006). An escort mechanism for cycling ofexport chaperones during flagellum assembly. Proc.Natl. Acad. Sci. U. S. A. 103, 17474–17479.

16. Auvray, F., Ozin, A. J., Claret, L. & Hughes, C. (2002).Intrinsic membrane targeting of the flagellar exportATPase FliI: interaction with acidic phospholipids andFliH. J. Mol. Biol. 318, 941–950.

17. Minamino, T., Gonzalez-Pedrajo, B., Yamaguchi, K.,Aizawa, S. I. & Macnab, R. M. (1999). FliK, the proteinresponsible for flagellar hook length control inSalmonella, is exported during hook assembly. Mol.Microbiol. 34, 295–304.

18. Bennett, J. C., Thomas, J., Fraser, G. M. & Hughes, C.(2001). Substrate complexes and domain organizationof the Salmonella flagellar export chaperones FlgN andFliT. Mol. Microbiol. 39, 781–791.

19. Fraser, G. M., Bennett, J. C. & Hughes, C. (1999).Substrate-specific binding of hook-associated proteinsby FlgN and FliT, putative chaperones for flagellumassembly. Mol. Microbiol. 32, 569–580.

20. Ehrbar, K., Winnen, B. & Hardt, W. D. (2006). Thechaperone binding domain of SopE inhibits transportvia flagellar and SPI-1 TTSS in the absence of InvB.Mol. Microbiol. 59, 248–264.

21. Lee, S. H. & Galan, J. E. (2004). Salmonella type IIIsecretion-associated chaperones confer secretion-pathway specificity. Mol. Microbiol. 51, 483–495.

22. Auvray, F., Thomas, J., Fraser, G. M. & Hughes, C.(2001). Flagellin polymerisation control by a cytosolicexport chaperone. J. Mol. Biol. 308, 221–229.

23. Weber-Sparenberg, C., Poplau, P., Brookman, H.,Rochon, M., Mockel, C., Nietschke, M. & Jung, H.(2006). Characterization of the type III export signal ofthe flagellar hook scaffolding protein FlgD of Escher-ichia coli. Arch. Microbiol. 186, 307–316.

882 Sorting of Early and Late Flagellar Subunits

24. Sorg, J. A., Blaylock, B. & Schneewind, O. (2006).Secretion signal recognition by YscN, the Yersinia typeIII secretion ATPase. Proc. Natl. Acad. Sci. U. S. A. 103,16490–16495.

25. Homma, M., Kutsukake, K., Hasebe, M., Iino, T. &Macnab, R. M. (1990). FlgB, FlgC, FlgF and FlgG. Afamily of structurally related proteins in the flagellarbasal body of Salmonella typhimurium. J. Mol. Biol. 211,465–477.

26. Homma, M., DeRosier, D. J. & Macnab, R. M. (1990).Flagellar hook and hook-associated proteins of Sal-

monella typhimurium and their relationship to otheraxial components of the flagellum. J. Mol. Biol. 213,819–832.

27. Datsenko, K. A. & Wanner, B. L. (2000). One-stepinactivation of chromosomal genes in Escherichia coliK-12 using PCR products. Proc. Natl. Acad. Sci. U. S. A.97, 6640–6645.

28. Ohnishi, K., Ohto, Y., Aizawa, S., Macnab, R. M. &Iino, T. (1994). FlgD is a scaffolding protein needed forflagellar hook assembly in Salmonella typhimurium.J. Bacteriol. 176, 2272–2281.