JOURNAL OF BACrERIOLOGY, Apr. 1994, p. 2272-2281 Vol. 176, No. 8

0021-9193/94/$04.00+0Copyright ©D 1994, American Society for Microbiology

FlgD Is a Scaffolding Protein Needed for Flagellar HookAssembly in Salmonella typhimurium

KOUHEI OHNISHI,1'2 YORIKO OHTO,1 SHIN-ICHI AIZAWA,3 ROBERT M. MACNAB,2*AND TETSUO IINO1

Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520-8114,2 andSchool ofHuman Sciences, Waseda University, Tokorozawa, Saitama 359,1 and Department of Biosciences,

Teikyo University, Utsonomiya, Tochigi-ken 320,3 Japan

Received 9 December 1993/Accepted 9 February 1994

FlgD is known to be absolutely required for hook assembly, yet it has not been detected in the matureflagellum. We have overproduced and purified FlgD and raised an antibody against it. By using this antibody,we have detected FlgD in substantial amounts in isolated basal bodies fromflgA,flgE,flgH,flgI,flgK, andfliKmutants, in much smaller amounts in those from the wild type andflgL,fli4,fliC,fliD, andfliE mutants, andnot at all in those fromflgB,flgD,flgG, andflgj mutants. In terms of the morphological assembly pathway, theseresults indicate that FlgD is first added to the structure when the rod is completed and is discarded when thehook, having reached its mature length, has the first of the hook-filament junction proteins, FlgK, added to itstip. Immunoelectron microscopy established that FlgD initially is located at the distal end of the rod andeventually is located at the distal end of the hook. Thus, it appears to act as a hook-capping protein to enableassembly of hook protein subunits, much as another flagellar protein, FliD, does for the flagellin subunits ofthe filament. However, whereas FliD is associated with the filament tip indefinitely, FlgD is only transientlyassociated with the hook tip; i.e., it acts as a scaffolding protein. When FlgD was added to the culture mediumof aflgD mutant, cells gained motility; thus, although the hook cap is normally added endogenously, it can beadded exogenously. When culture media were analyzed for the presence of hook protein, it was found only withthe flgD mutant and, in smaller amounts, the fliK (polyhook) mutant. Thus, although FlgD is needed forassembly of hook protein, it is not needed for its export.

The bacterial flagellum is a complicated structure composedof the basal body, the hook, and the filament (see, e.g.,reference 19), as well as more labile structures, such as themotor, switch, and export apparatus. Flagella, under thecontrol of the associated sensory apparatus, provide the cellwith the ability to move to favorable environments. Theflagellar basal body consists of subunits of at least eightdifferent proteins, which form two outer rings (the L and Prings), an inner ring (the MS ring), and the rod (Fig. 1). Thehook and the filament are homopolymers of hook protein andflagellin, respectively. The morphological pathway of flagellarformation is well characterized in both Escherichia coli andSalmonella typhimurium (13, 15, 26, 27) and is coordinated withflagellar gene expression (17). The flagellum is sequentiallyconstructed from simpler to more complex structures. At theearliest stage, the MS ring complex is formed from subunits ofthe FliF protein. It is thought that the flagellar switch and theflagellar export apparatus are then added (13, 15). Basal bodyassembly continues with formation of the rod and addition ofthe outer (P and L) rings. After the basal body is completed,the hook is assembled and finally polymerization of thefilament, the major external structure and the propeller for thecell, commences and continues indefinitely. Export of rod,hook, and filament subunits is believed to occur via a centralchannel in the nascent structure (19).Most of the genes for the flagellar hook-basal body complex

are clustered on the chromosome in flagellar region I (Fig. 2).

* Corresponding author. Mailing address: Department of MolecularBiophysics and Biochemistry 0734, Yale University, P.O. Box 208114,266 Whitney Ave., New Haven, CT 06520-8114. Phone: (203) 432-5590. Fax: (203) 432-9782. Electronic mail address: rmacnab@yalevm.ycc.yale.edu.

This region contains the genes for the proximal rod (flgB, flgC,and flgF), the distal rod (flgG), the L ring (figH), the P ring(figl), the hook (figE), and the hook-filament junction proteins(flgK and flgL) (4, 7, 8, 12). It also contains five other genes:flgN,flgM,flgA,flgD, and flgJ. Except for flgM, which codes foran inhibitor of the flagellum-specific sigma factor, uF (2, 23),and flgN, which has an unknown function (3), these genes arethought to be related to hook-basal body formation (26, 27).Assembly is blocked at the stage of hook protein polymeriza-tion inflgD mutants and at the stage of outer-ring formation inflgA mutants (15). Genes for the MS ring and various otherearlier stages of flagellar formation are found in region IIIb,which is remote from region I (40 versus 23 min) (19).

In a swimming cell, the flagella rotate at high speed (on theorder of 100 Hz; 18) and are under considerable torsional load(the load being even higher in the laboratory technique knownas tethering). The connection between the basal body rod andthe hook, and that between the hook and the filament, musttherefore be quite strong to avoid breakage. Between the hookand the filament, there exist two proteins (FlgK and FlgL),whose role may be to provide a stable junction (5); theseproteins are variously called hook-filament junction proteins orhook-associated proteins. Immediately after the hook is com-pleted, these proteins are assembled sequentially at the distalend of the hook: first FlgK, then FlgL, and finally a thirdprotein, FliD (6, 10). FliD acts as a capping protein thatpermits insertion of newly exported flagellin monomers at thefilament tip (5, 11); only after it has been added can filamentformation start. So, from the distal end of the rod, the order ofsubstructures is hook, FigK, FlgL, filament, and FliD. Incontrast, no junction protein between the rod and hook, orcapping protein for the hook, has been identified.

Although FlgD has not been detected in the purified flagel-

2272

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

FLAGELLAR HOOK ASSEMBLY IN S. TYPHIMURIUM 2273

Filament cap (FliD)

20 nmi-

figN + ?.M anti-( factor

A outer-ring assembly

Filament(FliC)

BCDEFGH

J IHook-filament junction

(FlgL)K (FlgK)

Hook (FIgE)

Distal rod(FIgG)

exterior

L ring (FIgH)/ P ring (Flgl)

mu outer membranepeptidoglycan layer

& periplasm-M~ cell membrane

MS ring (FliF) cytoplasm

FIG. 1. Illustration of the filament-hook-basal body complex of thebacterial flagellum of S. typhimurium, showing the various substruc-tures and the proteins from which they are constructed. (Othercomponents of the flagellum, such as the motor, switch, and exportapparatus, have been omitted for clarity.) The basal body consists of aninner MS ring, two outer P and L rings, and a rod (light cross-

hatching). Attached to the distal end of the rod is the hook (heavycross-hatching). The stippled boxes between the hook and filamentrepresent zones of the two hook-filament junction proteins; thestippled box at the tip of the filament is the filament cap. FlgD, thesubject of this study, is needed for hook assembly but is not locatedwithin the filament-hook-basal body complex. The relationship be-tween the flagellum and the layers of the gram-negative cell surface isindicated to the right.

lar structure (15), it seems a plausible candidate for a rod-hookjunction protein or hook-capping protein. In flgD mutants,flagellar formation ceases with completion of the basal body(26, 27); in other words, no hook assembly occurs. Thus, FlgDin some way enables hook protein subunits to add to the rodand so has been described as a rod modification protein. fig!

and flgH mutants, which construct an incomplete basal bodylacking one or both outer rings, respectively, also fail toassemble the hook structure, but in this case small amounts ofhook protein are found at the distal end of the rod (15, 26),proving that these mutants are capable of initiating hookpolymerization; the barrier to further polymerization appearsto be the outer membrane. In fact, when large amounts of hookprotein were synthesized from plasmid-encoded flgE, thesecells could overcome flgH or flg! mutations and form a hookand a filament (21). In contrast, flgD mutants have no hookprotein on the distal end of the rod (15) and cannot bypass theblock in hook assembly, even when large amounts of hookprotein are provided (21).

In this study, we tested for the presence of FlgD in partial

proximal rodproximal rodrod modificationhookproximal roddistal rodL ringP ring

IK hook-filament junctionL hook-filament junction

FIG. 2. Flagellar region I, located at 23 min on the S. typhimuriumchromosome. Gene designations for this region all begin with flg.Genes of special interest to this study are in boldface: flgD, whoseproduct is engaged in rod modification, a process that is needed beforehook assembly can occur; flgE, the structural gene for the hookprotein; flgG, the structural gene for the distal rod protein; and flgK,whose product lies at the distal end of the hook in the matureflagellum. The functions of other genes in the region, when known, arealso indicated. Operons are indicated by arrows and are not drawn toscale.

structures of various mutants and concluded that it is presentthroughout the entire process of hook assembly, after which itis discarded.

MATERIALS AND METHODS

Strains and plasmids. The S. typhimurium strains used arelisted in Table 1. All of the S. typhimurium flagellar mutantsused are derivatives of SJW1103, a flagellar wild-type strain(30). Strain MY2601 is an intragenic suppressor offliK mutantstrain SJW153 and has regained the capacity to assemble afilament (polyhook-filament phenotype). E. coli BL21(DE3)carries the T7 RNA polymerase gene (25). Plasmids pTTQ18(24) and pET-3 (25) are expression vectors.

Purification of FlgD. Cells of BL21(DE3) containing apET-3-based plasmid were cultured overnight in mediumcontaining 10 g of tryptone; 5 g of NaCl; 20 ml of a solutioncontaining 0.1 M Na2SO4, 1.2 M NH4H2PO4, 0.4 M Na2HPO4,and 2 M K2HPO4; 2 ml of 2 M glucose; 980 ml of distilledwater, and 50 jig of ampicillin ml- . A 250-ml volume of thismedium was inoculated with 1.25 ml of the overnight cultureand incubated at 37°C with shaking for 2.5 h. Incubation wascontinued for a further 2 h following addition of 1.5 ml of 100mM isopropyl-p-D-thiogalactopyranoside (IPTG); (final con-centration, 0.6 mM). Cells were harvested, washed, and sus-pended in 25 ml of 10 mM Tris-HCl (pH 7.5). The cellsuspension was sonicated twice for 60 s each time at 80 W(Branson W140D) and centrifuged at 8,000 x g for 10 min toremove cell debris. The supernatant was then centrifuged at100,000 x g for 1 h, and the supernatant was loaded on aQ-Sepharose FF column (2.5 by 6.1 cm; Pharmacia) equili-brated with 10 mM Tris-HCl (pH 7.5). Proteins were eluted in1.5-ml fractions with 100 ml of a linear 0 to 200 mM NaCl

VOL. 176, 1994

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

J. BACTERIOL.2274 OHNISHI ET AL.

TABLE 1. Strains used in this study

Genotype

Wild typeflgAflgBflgDflgDflgDflgDflgDflgDflgEflgFflgGflgHAlIflgJflgKflgLflhDflhDCfliAfliCfliDfliEfliHfliKfliKfliM

Function of mutated gene

Outer ring assemblyRod proteinUnknownUnknownUnknownUnknownUnknownUnknownHook proteinRod proteinRod proteinL-ring proteinP-ring proteinUnknownHook-filament junctionHook-filament junctionMaster regulatorMaster regulatorFlagellum-specific a factorFlagellinFilament capBasal body proteinExport apparatusControl of hook lengthControl of hook lengthMotor switch

Reference

302915291529152929292929152929292929292929292929292929

gradient and then 30 ml of 1 M NaCl. Fractions were moni-

tored by sodium dodecyl sulfate-polyacrylamide gel electro-

phoresis (SDS-PAGE) with Coomassie blue stain, and those

containing FlgD (which appeared at an apparent molecularmass of 27 kDa [the 27K protein]) were pooled. The pooledsolution was concentrated by dialysis against solid polyethyleneglycol and a 100-pld portion was separated on Protein PAK 300

(Waters) equilibrated with 50 mM sodium phosphate buffer

(pH 7.0)-100 mM sodium carbonate at a flow rate of 1 ml

min- and collected in 0.5-ml fractions. Fractions were mon-

itored by SDS-PAGE, and those containing FlgD were pooled.Polyclonal rabbit antibodies were raised against the FlgD 27K

protein band excised from SDS-PAGE gels. Immunoblottingwith anti-FlgD antibody was carried out as described in the

next section, except that the final detection employed Immu-

nostain HRP (Konica).Isolation of flagellar structures. Complete and incomplete

flagellar structures were isolated as described by Kubori et al.

(15), with minor modifications. For immunoblotting, 250 ml of

3% yeast extract medium was inoculated with 12.5 ml of a

Luria broth overnight culture and incubated with shaking at

37°C for 3 h. Following spheroplast induction with lysozyme,flagellar structures were solubilized from cell membranes with

Triton X-100, precipitated, and finally dissolved in 50 pul of

distilled water. This sample was mixed with 10 pul of 200 mMTris-HCl (pH 6.8)-6% SDS-60% glycerol-0.01% bromophe-nol blue and boiled for 3 min. A 10-pld portion of the samplewas separated on an SDS-12% polyacrylamide gel, and pro-

teins were transferred as previously described (22). Proteins

probed with the anti-FlgD or anti-FliF antibody were detected

by using the ECL system (Amersham).For electron microscopic observation, solubilized flagellar

structures were precipitated and dissolved in TET (10 mMTris-HCl [pH 8.0], 1 mM EDTA, 0.5% Triton X-100), to which

1.5 g of CsCl was added, and then the volume was adjusted to

5 ml. This solution was centrifuged at 40,000 x g for 14 h witha Beckman SW55 rotor. The sample was then removed fromthe tube in three equal portions (top, middle, and bottom) andcentrifuged at 100,000 x g for 1 h with a Beckman 7OTi rotor.

The precipitates were each dissolved in 50 RI of TET and usedfor microscopic observation.Assay for restoration of motility by exogenous FlgD. (i) In

liquid medium. A 0.1-ml portion of an overnight culture of theflgD mutant was inoculated into 5 ml of Luria broth andincubated with shaking at 37°C. After 2 h, 100-pul portions weretransferred to two new tubes, to each of which was added 50 pIof either a purified FlgD suspension (which had been passedthrough a 0.45-,um-pore-size filter) at a final concentration of10 ,ug ml 1 or distilled water. A portion of the cultures was

taken out every hour and observed by dark-field microscopy.Photographic motility tracks were taken with a 3-s exposuretime.

(ii) On semisolid plates. A 4-ml volume of motility agar (15g of Bacto Tryptone, 5 g of NaCl, and 2.5 g of agar dissolvedin 1 liter of distilled water) supplemented with various amountsof purified FlgD was poured into a 5-cm-diameter petri dish.Plates were dried at 30°C overnight, and a single colony of themutant to be tested was spotted in the center of each plate,which was then incubated at 30°C for 6 h.

Immunoelectron microscopy. Anti-FlgD antibody was puri-fied by ammonium sulfate precipitation. A 5-pl portion of theflagellar structure isolated as described above was mixed withthe antibody and incubated for 30 min at room temperature. ASO-pd volume of TET was added to the mixture, which was thencentrifuged at 100,000 x g for 1 h. The pellet was dissolved in10 [lI of TET. Portions were negatively stained with 2%(wt/vol) phosphotungstate (pH 6) on carbon films and ob-served in a JEM-100 electron microscope (JEOL Ltd., Tokyo,Japan).

Detection of hook protein in the medium. All of the strainsused were transformed either by aflgE-carrying pBR322-basedplasmid, pOH647 (21), to make detection of hook protein inthe medium easier, or by pBR322. A 2-ml sample of Luriabroth plus ampicillin was inoculated with 20 RI of an overnightculture. After 3 h of incubation at 37°C, 1 ml of culture was

removed and centrifuged for 15 min. (The rest of the cells wereused for measuring the optical density at 600 nm, which was

typically around 0.7 to 0.9.) A 200-pd volume of the superna-tant was removed and mixed with 40 pul of 200 mM Tris-HCl(pH 6.8)-6% SDS-60% glycerol-0.01% bromophenol blue,boiled for 3 min, and subjected to immunoblotting. Thesupernatant from a constant number of cells was loaded ineach lane on the basis of the following formula: volume added= 6 ,lI/(optical density at 600 nm). Samples were separated onSDS-12.5% PAGE and transferred to a polyvinylidene difluo-ride membrane as described by Towbin et al. (28). Themembrane was blocked for 1 h with 5% dry skim milk in 20mM Tris-HCl (pH 7.6)-150 mM NaCl-0.1% Tween 20. It wasnext incubated overnight with anti-FlgE antibody in 10 ml ofblocking solution and then for 1 h with goat anti-rabbitimmunoglobulin G-horseradish peroxidase (Bio-Rad) in 10 mlof blocking solution. Hook protein probed with antibody was

visualized by the chemiluminescence of the ECL system.

RESULTS

Construction of a FlgD-overproducing plasmid. To purifythe FlgD protein, we constructed a plasmid carryingflgD undercontrol of the T7 promoter (Fig. 3). A 1.55-kb AluI-PstIfragment containingflgD and the 5' half offlgE was cloned into

pUC18 to produce pKK1447. A 1.15-kb EcoRI-NruI fragment

Strain

SJW1103SJW1446SJW1525SJW156SJW157SJW158SJW197SJW11767SJW12295SJW1353SJW1444SJW1378SJW1469SJW1351SJW1435SJW2177SJW2172SJW1400SJW1368SJW1448SJW2536SJW2149SJW1371SJW1429SJW107MY2601SJW1407

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

FLAGELLAR HOOK ASSEMBLY IN S. TYPHIMURIUM 2275

N/S B N

N/S B H

K K NSB B

A/S K E e

FIG. 3. Physical map offlgD in several vectors. Plasmids pKK1447,pGD20, pGD30, and pEGD1 and pEGD2 are based on pUC18,pTTQ18, pBluescript II KS(+), and pET-3, respectively. Ptac and PT7represent the tac and T7 phage gene 10 promoters, respectively. Themultiple cloning sites from pUC18 (M) and pBluescript II KS(+)( ) are indicated. The extents of these regions are exaggerated.Restriction sites: A, Alul; B, BamHI; E, EcoRI; H, HindIII; K, KpnI;N, NruI; S, SmaI; P, PstI.

of pKK1447 containing flgD was recloned into the EcoRI andSmaI sites of pTTQ18 to give plasmid pGD20. The EcoRI-Hindlll fragment from pGD20 was recloned on pBluescript IIKS(+) to give plasmid pGD30, which had two BamHI sites.The BamHI-digested flgD-containing fragment of pGD30 wascloned into expression vector pET-3 in both orientations. Theplasmid with flgD in the same orientation as the T7 promoterwas termed pEGD1, and the other was termed pEGD2.Plasmids pET-3, pEGD1, and pEGD2, were introduced intoE. coli BL21(DE3). When IPTG-induced proteins in thesetransformants were separated on SDS-polyacrylamide gels, aprotein with an apparent molecular mass of 27 kDa (the 27Kprotein) was seen in cells transformed with pEGD1 (data notshown); this molecular mass is in reasonable agreement withthe value of 23,961 Da deduced from the nucleotide sequence(16). In cells transformed with pEGD2 or pET-3, this 27Kprotein was not seen. From these results it seemed likely thatthe 27K protein was the flgD gene product.

Purification of FlgD and preparation of anti-FlgD antibody.We used BL21(DE3) cells transformed with pEGD1 to over-produce the 27K protein and purify it. Cells induced withIPTG were harvested and sonicated. The sonicated fractionwas centrifuged at low speed, and the supernatant was furthercentrifuged at high speed to separate the soluble (periplasmicand cytoplasmic) fraction from the membrane fraction. Whenall three fractions (low-speed pellet and soluble and membranefractions) were analyzed on SDS-polyacrylamide gels, the 27Kprotein was detected mainly in the soluble fraction (Fig. 4A,lane 1). Proteins in this fraction were loaded onto an ion-exchange column and separated, and peak fractions containingthe 27K protein were pooled (Fig. 4A, lane 2), concentrated,and separated by gel filtration. The 27K protein was almost100% pure in the peak fractions (Fig. 4A, lane 3).The material from the final purification step was separated

by SDS-PAGE and transferred to a membrane, and thesequence of the first 20 N-terminal amino acids of the 27Kprotein was determined. Except for loss of the initial methio-nine, which presumably had been removed posttranslationally,the sequence was identical to the amino acid sequence de-duced from the nucleotide sequence of flgD (16). Thus, the27K protein was confirmed to be FlgD.FlgD separated by SDS-PAGE was excised from gels and

used to prepare anti-FlgD serum. The purified anti-FlgDantibody was examined for effectiveness in recognizing FlgD.Proteins in the pooled fractions from each purification stepwere separated by SDS-PAGE, transferred to polyvinylidene

3A) M 1 2 3 B) M 1 2200:f),w .; 106-

80-66-_5- 49.5-

36- - 32.5-

31- 27.5-

18.5-

FIG. 4. Purification of FlgD. (A) A sample at each purification stepwas analyzed by SDS-12.5% PAGE. Cells of BL21(DE3)/pEGD1induced by IPTG were harvested, sonicated, and centrifuged at lowspeed, and the supernatant was centrifuged at high speed. Lanes: 1,supernatant from the high-speed centrifugation; 2, Q-Sepharose FFfraction; 3, protein PAK 300 fraction. Note the major band at anapparent molecular mass of 27 kDa. The faint bands below the majorband in lanes 2 and 3 are likely to be degradation products of it. LaneM contained molecular mass standards (sizes are in kilodaltons). (B)The same samples as in panel A were analyzed by SDS-PAGE andtransferred to a polyvinylidene difluoride membrane. The membranewas incubated with anti-FlgD antibody and then with goat anti-rabbitimmunoglobulin G-horseradish peroxidase and visualized by immu-nostaining. Lane M contained prestained molecular mass markers(sizes are in kilodaltons).

difluoride membranes, and reacted with antibody. In all frac-tions, one major band with an apparent molecular mass of 27kDa reacted with the antibody (Fig. 4B), showing that thelatter could be used for immunodetection of FlgD.Assay for FlgD activity. We wished to know whether purifiedFlgD had physiological activity. Pooled fractions from theion-exchange chromatography or gel filtration step werepassed through a 0.45-,um-pore-size membrane filter to re-move any contaminating bacteria and added to a final concen-tration of 10 ,ig ml-l to a liquid culture of a flgD mutant.Within an hour, a few cells had become motile, and within 3 to5 h, almost all had done so (Fig. 5, right panels); the swimmingpatterns of these cells were indistinguishable from those ofwild-type cells. Pooled fractions from the ion-exchange chro-matography and gel filtration steps were equally effective inrestoring motility. In a control in which no FlgD was added tothe medium, flgD mutant cells remained immotile indefinitely(Fig. 5, left panels).When the flgD mutant was spotted on motility agar platescontaining purified FlgD, it formed swarms. The diameters of

the swarms increased in proportion to the amount of FlgDadded (Fig. 6, left panels). With the flgI mutant (whichconstructs an incomplete basal body lacking the outer ring andfails to form a hook or filament), no swarms were observed,even with high concentrations of added FlgD (Fig. 6, rightpanels). These results indicate that purified FlgD can actexternally from the cell to enable hook protein polymerizationand that this activity is limited to flgD mutants.

Detection of FlgD in flagellar precursor structures. Incom-plete flagellar structures from various mutants were isolated,separated by SDS-PAGE, and transferred to a membrane,where they were probed by antibodies against FlgD and FliF(the structural protein of the simplest detectable flagellarstructure, the MS ring) (Fig. 7). A band corresponding to FliF,

A 100 bpK K N i, pI I I IflgD 0 figE - --1

pKK1447 Emzz/E K A/S

BE K A/SpGD20 3

pGD30 _-

pEGD1 B EKA/SpEGD2

K K

VOL. 176, 1994

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

J. BACrERIOL.2276 OHNISHI ET AL.

-FlgD +FlgD

Ohr

1 hr

3hr

5hr

FIG. 5. Effect of exogenous FlgD on the motility of a

A 50-,ul volume of distilled water (left panels) or a Isuspension (right panels) was added to 100 ,ul of an e;SJW156 (flgD) culture. Portions were withdrawn at thaddition indicated and observed by dark-field microsctracks were generated by using a 3-s exposure time.

with an apparent molecular mass of 65 kDa, was o

all flagellar mutants except an flhD mutant (whiciin the flagellar master operon and does not syi

flagellar proteins). FlgD, in contrast, was detectamounts only in flg4, flgE, flgH, flgI, flgK, and flilwas detected in small amounts in flgL, fliA, fliC, Jmutants, in extremely small amounts in the wild tat all in flgB, flgD, flgG, flgJ, and flhD mutants. lwas present as a major component only at earl

flagellar formation, from the stage of the basalouter rings through the hook-basal body stage jufirst hook-filament junction protein, FlgK, hasThese results suggest that FlgD is a scaffoldinflagellar hook formation. Interestingly, the amc

present in the fliK (polyhook) mutant was substan(by a factor of around 3) than in a flgK mutant.

Localization of FlgD in the hook-basal bo4performed electron microscopy and immunoelescopy to determine where FlgD is located wiprecursor structures (Fig. 8).

Hook-basal bodies from a flgK mutant, SJVcontain FlgD in their flagellar structure (Fig. 7)and examined. As well as intact hook-basal bcprecursors called rivets, which lack the outer rinwere observed (cf. reference 15). In the absencehook-basal bodies and rivets were often seen iIassociation (either parallel or antiparallel) via th

4 3 4 5

FIG. 6. Motility recovery on semisolid plates as a result of additionof exogenous FIgD. A single colony of SJW156 (flgD) (left) orSJW1351 (flgI) (right) was spotted onto semisolid tryptone agar plates1 to 5 supplemented with purified FlgD at concentrations of 1, 0.25,0.06, 0.01, and 0 p.g ml- 1, respectively. The plates were incubated at30°C for 6 h.

rings (Fig. 8A). Such associations, which have been reportedpreviously (1, 15), are believed to result from nonspecifichydrophobic interactions.

In the presence of anti-FlgD antibody, several different typesE of association were noted (Fig. 8B): between the hook tips of

hook-basal bodies, between the hook tip of a hook-basal bodyand the rod tip of a rivet, and between the rod tips of rivets.These results demonstrate that FlgD exists at the tip of thehook or, in the case of particles lacking a hook, at the tip of therod. Particles isolated from a fliK (polyhook) strain, SJW107,also aggregated tip to tip in the presence of anti-FlgD antibody

501Am (Fig. 8C), consistent with the fact that they retain FlgD (Fig. 7).When we prepared hook-basal bodies from a flgL mutant,

ifgD mutant. SJW2172, where FlgK is assembled onto the tip of the hook (5)purified FlgD and FlgD is no longer present as a major component (Fig. 7),

arly-log-ph se anti-FlgD antibody did not cause the tip-to-tip aggregates that

copy. Motility were observed in the flgK or fliK mutant. Only ring-to-ring*pyMotility aggregates, similar to those found in hook-basal body prepa-

rations without antibody, were observed (Fig. 8D). This indi-cates that FlgD at the tip of hook is displaced by FlgK prior to

tbserved with filament formation.

h is defective We attempted to purify partial structures containing FlgD to

nthesize any estimate its stoichiometry within the basal body, but the

ted in major samples were too heterogeneous. We suspect that by analogyY mutants. It with the filament cap, the number of subunits will be quite

fliD, and fliE small, perhaps around 5 to 10 (11).type, and not Secretion of hook protein (FlgE) in flagellar mutants.

rhat is, FlgD Although flgD mutants cannot assemble a hook, the hook

lier stages of protein (FlgE) might nonetheless be exported and secreted

body lacking into the medium. To test this, we carried out SDS-PAGE of

LSt before the the culture medium supernatants of various mutants (trans-been added..g protein inpunt of FlgD ",P_1>

'1?4_;

itially greater __

dy. We nextctron micro-thin flagellar FlgD -

V2177, which FIG. 7. Detection of FlgD in flagellar precursors. Complete orwere isolated incomplete hook-basal bodies were purified from mutants defective inides, simpler the genes shown above the lanes (for strain numbers, see Table 1) and

lgS and hook, then subjected to SDS-PAGE and immunoblotting with antibodiesof antibody, against FlgD and (as a control) FliF, the basal-body MS-ring protein.

n side-by-side st, purified FlgD protein standard; wt, wild type. In lane wt, a faint

eir L and MS band is visible on the original autoradiogram but not in this figure.

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

l - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-

I ~

D

FIG. 8. Localization of FlgD within flagellar precursors. Shown are electron micrographs and immunoelectron micrographs of purifiedhook-basal bodies prepared from SJW2177 (flgK) (A and B), SJW1O7 (fliK) (C), and SJW2172 (flgL) (D). Panel A shows hook-basal bodies in theabsence of anti-FlgD antibody. For the other panels, hook-basal bodies were incubated with anti-FlgD antibody for 30 min at room temperature,pelleted at 100,000 x g, and dissolved in TET. The arrow in panel C indicates the aggregation point of the polyhook tips. Bar, 50 nm.2277

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

2278 OHNISHI ET AL.

formed with a plasmid containing flgE to make it easier todetect hook protein in the medium), transferred them to amembrane, and probed them with anti-FigE antibody (Fig. 9,top). With anflhDC mutant, which does not make any flagellarproteins, no hook protein was detected- in the medium eventhough it was being overproduced in the cells, establishing thatcell lysis was negligible in these experiments. In the medium ofSJW157 (flgD), hook protein was detected in large amounts;the amounts increased with cell culture density (data notshown). Other flgD mutants (SJW156, SJW158 [Fig. 9, bot-tom], SJW197, SJW11767, and SJW12265 [data not shown])also secreted hook protein into the medium, although theamount varied between alleles. With several of these mutants,the amounts of hook protein secreted as a result of chromo-somal expression alone (i.e., in the absence of the figE-containing plasmid) were sufficient to be detectable (e.g., lane4 in Fig. 9, bottom). We conclude that FlgD is not needed forhook protein export.Under overproducing conditions, small amounts of hook

protein were detected in the medium of other strains that docomplete the basal body, such as the wild type (visible on theoriginal film but not the final print) and the fliK mutant (Fig.9). With all of the mutant strains tested that fail to construct acomplete basal body (those defective in flhD, fliE, fliH, fliM,flgA,flgB,flgC,flgF,flgG,flgH,flgI, orflgl [data not shown]), wefailed to detect hook protein in the medium, even underconditions of overproduction.

DISCUSSION

The process of flagellar assembly is a complex one with manyinteresting aspects. Several of these concern assembly of thehook. (i) Since this is a structure external to the cell, itscomponent subunits need to be exported. (ii) It is assembled toa defined length. (iii) The process takes place at a definedpoint in the assembly pathway, after formation of the basalbody rod and before that of the filament.FlgD is a scaffolding protein. It has been known since the

pioneering studies of Suzuki and coworkers (26, 27) that theFlgD protein is absolutely required for formation of theflagellar hook, with the rod tip in flgD mutants exhibiting aclean appearance that indicates total failure to initiate hookassembly. The FlgD-dependent event that enables hook pro-tein subunits to attach to the rod has been called rod modifi-cation, but the nature of the modification and the specific roleof FlgD in the process have been obscure.As the biochemistry of the flagellum became better under-

stood, it was found that FlgD was not a structural component,at least not of the mature flagellum (15). If not that of astructural component, what might its role be? Possibilitiesinclude export of the hook protein and the actual assemblyprocess itself. This study ruled out the former possibility, sinceflgD mutants export the hook protein in abundance (Fig. 9).

Control of assembly is clearly an important feature for thefilament, at the tip of which a capping protein called FliDresides and permits new flagellin subunits to assemble ratherthan be lost to the medium. FliD remains at the tip indefinitely,but this is presumably because filament elongation is a processthat goes on indefinitely and is the last event in flagellarassembly. The same is not true of the hook, whose assemblyproceeds to a well-defined length and then stops and issuperseded by filament assembly. We realized that this mightexplain why the mature flagellar structure lacks FlgD-per-haps it was being used in the assembly process and thendiscarded. In the present study, we have established that this isindeed the case.

FIgE _(hook protein)

FIgE(hook protein)

f/gD fliK1 2 3 4 5 6 7 8 9 101112

P U PUP c P P P P PP

FIG. 9. Detection of the hook protein, FlgE, in culture superna-tants of various strains by SDS-PAGE and immunoblotting withantibody against the hook protein, FlgE. Top, supernatants frommutants defective in the genes shown (see Table 1 for strain numbers)transformed with flgE-containing plasmid pOH647 to make FlgEdetection easier. wt, wild type. Bottom, supernatants from theflgD andfliK mutant strains indicated; all carry the prefix SJW (e.g., SJW156),except MY2601. P (plasmid flgE expression), cells transformed withpOH647; C (chromosomal flgE expression), cells transformed withpBR322.

The role of FlgD in flagellar morphogenesis is illustrated inFig. 10. FlgD is first incorporated into the nascent structurewhen the rod has been completed by assembly of its distalportion, made of FlgG subunits. At that stage, FlgD is locatedat the rod tip. It remains associated with the nascent hook untilthe latter has reached its mature length, being at that stagelocated at the hook tip. It seems almost certain (by analogywith the filament cap) that the hook cap resides at the tipthroughout the process and that the point of assembly of newhook protein subunits is at the distal end, just under the cap.The next step following completion of the hook is replacementof the hook cap by the first hook-filament junction zone, madeof FlgK (5). The process is apparently one of active displace-ment, since FlgD does not dissociate in the absence of FlgK(Fig. 7).

Subsequently, FlgL and FliD add to the tip of the growingstructure to make the second hook-filament junction zone andfilament cap, and finally the filament assembles by insertion offlagellin monomers just under the cap.The data obtained forfliA mutants at first do not appear to

fit in with the above scheme. FliA is a flagellum-specific cufactor (22) for expression of late flagellar genes, includingthose for hook-filament junction protein FlgK (17). Thissuggests that a fliA mutant should behave like a flgK mutant;i.e., the hook cap should be retained and assembly should stopat this point. FlgD was indeed detected, but the amounts werefar smaller than with a flgK mutant (Fig. 7). This result couldbe explained if the cap was being displaced, inefficiently, as aresult of low-level, FliA-independent expression of flgK. Thisinterpretation is consistent with the finding of Homma et al.(9) that the hook-associated proteins, including FlgK, can bedetected in hook-basal bodies from fliA mutants.Now that the location and fate of FlgD are known, we can

J. BACTERIOL.

.:r_l.1- * .* -. iO ..- I* -::1- .a_ --I.;-P - ---.,o d:). tp .P- tp0 kC% lw ;'Y-'7;tP C) iO .IN 0 .. .---

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

FLAGELLAR HOOK ASSEMBLY IN S. TYPHIMURIUM 2279

;"w. OM* PG

A CM

0FIgBFlgCFlgFFliE?

02

%F/i FigG n , FIgDFI

FliE? XiXon

- * AOM%J .- .,fIgAFlgl

MS RING/SWITCH

FIgH lw~~- RN

L RING

PROXIMAL R(

FIgE ~Str V//MMS

DISTAL ROD ROD/HOOK CAP(rod modification)

P RING

.. . FlgD~~ ~ ~ ~ ~ ~ ~ ~z) 6

WWII

X X - offZ111 fliA

fliK t""' FlgK t :s%. FIgL FliD FliC-%M VWNo%J IIb

NASCENT HOOK FULL-LENGTH HOOK 1st HOOK-FILAMENTJUNCTION PROTEIN

(displacement of hook cap) FILAMENT

FIG. 10. Role of FlgD in flagellar morphogenesis. Where a given protein is incorporated into the structure, the protein symbol is used (e.g.,FlgB); where a given gene is needed but its product is not believed to be incorporated, the gene symbol is used (i.e.,flgA,fliK, andfliA). Pre-existingstructural elements at any given stage are indicated by stippling. The newly added structural element at each stage is indicated in white. FlgD itselfis indicated in black throughout. The part of the overall pathway shown here begins at the point where the MS ring and the switch complex arealready assembled. In stage 1, the proximal rod is constructed from subunits of FlgB, FlgC, and FlgF, which are exported by the flagellum-specificexport pathway. In stage 2, the distal rod is constructed from FlgG. FliE is probably assembled in either stage 1 or stage 2, but its location is notknown. In stage 3, FlgD attaches to the distal end of the rod, forming a cap. In stage 4, the periplasmic P ring, made of FlgI, is added in aflgA-dependent process. In stage 5, the outer membrane L ring, made of FlgH, is added. In stage 6, hook protein (FlgE) subunits are ihsertedbetween the distal end of the rod and the FlgD cap; this event can initiate before assembly of the P and L rings but cannot proceed to a significantextent (see text). In stage 7, hook growth continues with insertion of FlgE subunits underneath the FlgD cap; the process is under control of thefliK gene and stops when the hook has reached its mature length. In stage 8, the FlgD cap is displaced by the first hook-filament junction protein,FlgK; FlgK is the first protein in the pathway to be synthesized by using flagellum-specific a factor FliA. In stage 9, the second hook-filamentjunction protein, FlgL, and the filament-capping protein, FliD, are added successively and flagellin subunits (FliC) are added distally by insertionunder the cap. With the exception of the outer-ring proteins FlgH and FlgI, all of the proteins assembled in the part of the pathway shown areexported by a flagellum-specific pathway that utilizes a central channel in the nascent structure. CM, cell membrane; PG, peptidoglycan layer andperiplasm; OM, outer membrane.

dismiss the possibility that it is a rod-hook junction proteinwhich simply escaped detection. Thus, the rod and hookappear to be capable of making a mechanically strong junctionwithout assistance from other proteins, while the same is nottrue of the junction between the hook and the filament. Thereason for this difference is not clear, especially since therod-hook junction has to withstand an even greater torsionaland lateral load than the hook-filament junction.Why does FlgD get discarded? We suspect that its properties

as a hook-capping protein would not be compatible with thoserequired for a hook-filament junction protein. In other words,it performs one function and is replaced by proteins (FlgK andFlgL) that perform a distinct function, that of creating aneffective junction between the hook and the filament (cf.reference 14).

Detection of minor amounts of FlgD in certain mutantclasses. The scheme shown in Fig. 10 is consistent with theresults obtained with those mutants in which FlgD was foundin major amounts in isolated basal bodies. However, we alsofound FlgD in minor amounts in the basal bodies from other(flgL, fliC, and fliD) mutants, where it would be predicted tohave been displaced by FlgK.How can these results be explained? A given mutation

defines the most advanced precursor structure that can besynthesized. Less advanced precursors will also be present,however, simply because assembly is still in progress (examples

of this sort can be seen in Fig. 8). Thus, a fliC mutant has as itsmost advanced precursor a hook-basal body with FlgD com-pletely displaced and with all three hook-associated proteinspresent at the tip of the hook; however, a nascent structurecould still be in the process of hook assembly, for example, andwould still have FlgD at its tip. This provides a simpleexplanation for the detection of small amounts of FlgD in thehook-basal body preparations from such mutants. In wild-typecells, the process of filament elongation probably representssuch a major process that the fraction of organelles still in thepreceding stages is very small, and hence only trace amounts ofFlgD are detected.

fliE mutants represent a more complicated situation. FliE isa basal-body component whose detailed location and functionare not known (20). In studies of isolated precursor structuresfrom mutants, Kubori et al. (15) found that FliE and the fourrod proteins (FlgB, FlgC, FlgF, and FlgG) were either allpresent or all absent; i.e., a defect in any one resulted in failureto detect the others, suggesting that intermediate levels of rodassembly are metastable and that FliE also plays a role in rodstability. They also found that all five proteins were present inprecursors from aflgD mutant, indicating that they precede itsaddition. (They were unable to comment on the presence orabsence of FlgD itself in any of the precursors they studied,because the protein had not been characterized at that time.)Consistent with this, we failed to detect FlgD in precursors

VOL. 176, 1994

OD

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

J. BACrERIOL.2280 OHNISHI ET AL.

from flgB or flgG mutants (flgC and flgF mutants were not

tested). However, we detected FlgD (albeit at low levels) inbasal bodies from fliE mutants. Perhaps the actual site forFlgD binding is FlgG, but the rod is only marginally stable inthe absence of FliE. Clearly, FliE itself needs to be studiedfurther before its relationship to FlgD can be understood.Hook length control. Does FlgD contribute to the mecha-

nism that controls the length of the hook, for example, as a

ruler? We do not know the answer, although several facts tendto argue against such a role. (i) Abnormal length is not

associated with FlgD but with another protein, FliK. (ii) Onemight expect the processes of export and length control to belinked, yet FlgD is not necessary for hook protein export.Finally, the location of the protein at the tip of the structuredoes not seem appropriate for a molecule that controls thestructure's length. However, the fact that we detected the

protein at the hook tip does not preclude the possibility that

part of it is located within the length of the hook. In theextended conformation, only about 180 amino acids would beneeded to span the 55-nm length of a mature hook (14). FlgD,with 231 amino acids (16), would be long enough to performthis role and still leave a significant amount of the proteinsequence at the tip.There are two experimental observations that may indicate

that the role of FlgD is more complex than we imagine. (i) Thepolyhooks that are assembled in fliK mutants contain more

FlgD than do the normal-length hooks offlgK mutants (Fig. 7),a result that is not easy to explain if FlgD forms a well-definedcap. (ii) The N-terminal sequence of FlgD suffices to supportassembly of hook protein (16), suggesting that the C-terminalportion of the molecule has another role, such as lengthcontrol.FlgD as a member of the axial family of exported flagellar

proteins. The axial family of proteins (rod proteins, hook

protein, hook-filament junction proteins, flagellin, and fila-

ment-capping protein) have been analyzed from the point of

view of recognition for export by the flagellum-specific path-way (4). Sequence similarities among the proteins were seen,but it was realized that they were likely to be, at least in part,reflections of structural similarities among the rod, hook, and

filament. The filament-capping protein, FliD, showed the least

similarity to the rest of the family, probably because of its

unique location at the tip. FlgD must now be included as a

member of the family of exported proteins. The sequence of

FlgD (16; accession number D25293) shows no significantsimilarity to that of FliD, despite their related functions, nor

does it show similarity to the distal rod protein, FlgG or the

hook protein, FlgE. This result reinforces the opinion ex-

pressed previously (4) that the signal for flagellum-specificexport is not readily apparent from the primary structure and

presumably derives from higher-order motifs.

ACKNOWLEDGMENTS

We thank K. Kutsukake for the gift of plasmid pKK1447 and for

provision of unpublished information regarding the flgD sequence and

the phenotype of certain flgD mutants, S. Yamaguchi for provision of

strains, and S. Lee for assistance with the assays of hook proteinexport.

This work was supported by a Grant-in-Aid for Scientific Research

from the Ministry of Education, Science, and Culture, Japan; a

research grant from Waseda University (to K.O.); and USPHS grantA112202 (to R.M.M.).

REFERENCES

1. Aizawa, S.-I., G. E. Dean, C. J. Jones, R. M. Macnab, and S.

Yamaguchi. 1985. Purification and characterization of the flagellarhook-basal body complex of Salmonella typhimunium. J. Bacteriol.161:836-849.

2. Gillen, K. L., and K. T. Hughes. 1991. Molecular characterizationofflgM, a gene encoding a negative regulator of flagellin synthesisin Salmonella typhimurium. J. Bacteriol. 173:6453-6459.

3. Gillen, K. L., and K. T. Hughes. 1993. Transcription from two

promoters and autoregulation contribute to the control of expres-sion of the Salmonella typhimurium flagellar regulatory gene flgM.J. Bacteriol. 175:7006-7015.

4. Homma, M., D. J. DeRosier, and R. M. Macnab. 1990. Flagellarhook and hook-associated proteins of Salmonella typhimurium andtheir relationship to other axial components of the flagellum. J.Mol. Biol. 213:819-832.

5. Homma, M., and T. lino. 1985. Locations of hook-associatedproteins in flagellar structures of Salmonella typhimurium. J.Bacteriol. 162:183-189.

6. Homma, M., T. lino, K. Kutsukake, and S. Yamaguchi. 1986. Invitro reconstitution of flagellar filaments onto hooks of filament-less mutants of Salmonella typhimunium by addition of hook-associated proteins. Proc. Natl. Acad. Sci. USA 83:6169-6173.

7. Homma, M., K. Kutsukake, M. Hasebe, T. Iino, and R. M.Macnab. 1990. FlgB, FlgC, FlgF and FlgG. A family of structurallyrelated proteins in the flagellar basal body of Salmonella typhi-murium. J. Mol. Biol. 211:465-477.

8. Homma, M., K. Kutsukake, and T. lino. 1985. Structural genes forflagellar hook-associated proteins in Salmonella typhimurium. J.Bacteriol. 163:464-471.

9. Homma, M., K. Kutsukake, T. lino, and S. Yamaguchi. 1984.Hook-associated proteins essential for flagellar filament formationin Salmonella typhimunium. J. Bacteriol. 157:100-108.

10. Ikeda, T., S. Asakura, and R. Kamiya. 1989. Total reconstitutionof Salmonella flagellar filaments from hook and purified flagellinand hook-associated proteins in vitro. J. Mol. Biol. 209:109-114.

11. Ikeda, T., M. Homma, T. lino, S. Asakura, and R. Kamiya. 1987.Localization and stoichiometry of hook-associated proteins withinSalmonella typhimurium flagella. J. Bacteriol. 169:1168-1173.

12. Jones, C. J., M. Homma, and R. M. Macnab. 1989. L-, P-, andM-ring proteins of the flagellar basal body of Salmonella typhi-murium: gene sequences and deduced protein sequences. J. Bac-teriol. 171:3890-3900.

13. Jones, C. J., and R. M. Macnab. 1990. Flagellar assembly inSalmonella typhimurium: analysis with temperature-sensitive mu-

tants. J. Bacteriol. 172:1327-1339.14. Jones, C. J., R. M. Macnab, H. Okino, and S.-I. Aizawa. 1990.

Stoichiometric analysis of the flagellar hook-(basal-body) complexof Salmonella typhimurium. J. Mol. Biol. 212:377-387.

15. Kubori, T., N. Shimamoto, S. Yamaguchi, K. Namba, and S.-I.Aizawa. 1992. Morphological pathway of flagellar assembly inSalmonella typhimurium. J. Mol. Biol. 226:433-446.

16. Kutsukake, K. Personal communication.17. Kutsukake, K., Y. Ohya, and T. lino. 1990. Transcriptional analysis

of the flagellar regulon of Salmonella typhimurium. J. Bacteriol.172:741-747.

18. Lowe, G., M. Meister, and H. C. Berg. 1987. Rapid rotation offlagellar bundles in swimming bacteria. Nature (London) 325:637-640.

19. Macnab, R. M. 1992. Genetics and biogenesis of bacterial flagella.Annu. Rev. Genet. 26:129-156.

20. Muller, V., C. J. Jones, I. Kawagishi, S.-I. Aizawa, and R. M.Macnab. 1992. Characterization of the fliE genes of Escherichiacoli and Salmonella typhimurium and identification of the FliEprotein as a component of the flagellar hook-basal body complex.J. Bacteriol. 174:2298-2304.

21. Ohnishi, K., M. Homma, K. Kutsukake, and T. lino. 1987.Formation of flagella lacking outer rings by flaM, flaU, and flaYmutants of Escherichia coli. J. Bacteriol. 169:1485-1488.

22. Ohnishi, K., K. Kutsukake, H. Suzuki, and T. lino. 1990. GenefliAencodes an alternative sigma factor specific for flagellar operons inSalmonella typhimurium. Mol. Gen. Genet. 221:139-147.

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

FLAGELLAR HOOK ASSEMBLY IN S. TYPHIMURIUM 2281

23. Ohnishi, K., K. Kutsukake, H. Suzuki, and T. lino. 1992. A noveltranscriptional regulation mechanism in the flagellar regulon ofSalmonella typhimunium: an anti-sigma factor inhibits the activityof the flagellum-specific sigma factor, (F. Mol. Microbiol. 6:3149-3157.

24. Stark, M. J. R. 1987. Multicopy expression vectors carrying the lacrepressor gene for regulated high-level expression of genes inEscherichia coli. Gene 51:255-267.

25. Studier, F. W., A. H. Rosenberg, J. J. Dunn, and J. W. Dubendorff.1990. Use of T7 RNA polymerase to direct expression of clonedgenes. Methods Enzymol. 185:60-89.

26. Suzuki, T., T. Iino, T. Horiguchi, and S. Yamaguchi. 1978.

Incomplete flagellar structures in nonflagellate mutants of Salmo-nella typhimurium. J. Bacteriol. 133:904-915.

27. Suzuki, T., and Y. Komeda. 1981. Incomplete flagellar structuresin Escherichia coli mutants. J. Bacteriol. 145:1036-1041.

28. Towbin, H., T. Staehelin, and J. Gordon. 1979. Electrophoretictransfer of proteins from polyacrylamide gels to nitrocellulosesheets: procedure and some applications. Proc. Natl. Acad. Sci.USA 76:4350-4354.

29. Yamaguchi, S. Unpublished data.30. Yamaguchi, S., H. Fujita, K. Sugata, T. Taira, and T. lino. 1984.

Genetic analysis of H2, the structural gene for phase-2 flagellin inSalmonella. J. Gen. Microbiol. 130:255-265.

VOL. 176, 1994

on February 25, 2021 by guest

http://jb.asm.org/

Dow

nloaded from