Characterization of flagella genes of Agrobacteriumtumefaciens , and the effect of a bald strain on virulence

Olga Chesnokova, John B. Coutinho, Imran H. Khan, †

Maurice S. Mikhail ‡ and Clarence I. Kado *

Davis Crown Gall Group, University of California, Davis,California 95616, USA.

Summary

Agrobacterium tumefaciens produces flagella that arearranged circumthecally near one end of the bacilli-form cell. The flagella are required for motility tofacilitate reaching the root surface, and possibly aidin orientating the bacterial cells at various sitesfor infection. We have identified three flagella genesdesignated flaA , flaB , and flaC. Mutations in flaA , flaBand flaC result in abberant swimming behaviour.Electron microscopic examination of these mutantsrevealed the defective flagella. A non-motile, baldmutant strain was generated by deleting all three flagenes. Nucleotide sequencing of flaA , flaB , and flaCshowed that they have a potential coding capacityfor polypeptides of 307, 321, and 314 amino acid resi-dues, respectively. The predicted amino acid sequencesof the A. tumefaciens FlaA and FlaB proteins are simi-lar (66% average identity) to the FlaA and FlaB proteinsencoded by flaA and flaB genes, respectively, in Rhizo-bium meliloti . There was no counterpart FlaC proteinreported in R. meliloti , but the A. tumefaciens FlaC issimilar in amino acid sequence to the R. meliloti FlaA(59.8% identity) and FlaB (66.7% identity). Distinctfrom FlaA and FlaB of R. meliloti is the absence of his-tidine and cysteine residues and their shorter length(by 88 amino acid residues fewer than FlaA and FlaBof R. meliloti ). The transcriptional start sites of eachfla gene determined by primer extension revealed con-sensus-sequence boxes representing potential bind-ing sites for s28 RNA polymerase (RNAP) upstream ofthe transcriptional start of each fla gene. Besidesthe potential s28-binding site upstream of flaC, alsopresent are additional putative conserved sequences,GC at ¹11 and GG at ¹21 from the transcriptionalstart, that resemble potential binding motifs for s54.

Because the s54 promoter is associated with genesregulated by physiological changes in various bac-teria, the flaC gene might be similarly regulated inresponse to A. tumefaciens responding to host plantstimuli. Virulence studies showed that the bald strainwas consistently reduced in virulence below that ofthe parental wild-type strain by at least 38%. The dif-ference is statistically significant and suggests thatthe flagella may play a role in facilitating virulence.

Introduction

The tumour-inducing plant pathogen Agrobacterium tume-faciens mediates the transfer of oncogenes carried on asegment of a resident tumour-inducing Ti plasmid fromthe bacterium to host plant cells, culminating in the forma-tion of non-self-limiting tumours. The initial infection pro-cess in soil requires the nearby presence of plant rootstoward which A. tumefaciens cells swim and becomeattached. For motility, A. tumefaciens produces flagellathat are circumthecally arranged, near one end ratherthan ringing the middle portion of the bacilliform cell, andthey are not peritrichously situated (Tanaka, 1985; Kado,1992). This type of flagellar arrangement might play arole in facilitating bacterial chemotaxis and expedite theinfection process because cells are frequently attachedto plant cells in a polar fashion (Smith and Hindley,1978; M. Hawes, unpublished). Motility, chemotaxis, andattachment by A. tumefaciens all seem to play criticalroles in the early infection process (reviewed in VandeBroek and Vanderleyden, 1995).

Thus, several laboratories have explored the potentialrole of motility and chemotaxis on Agrobacterium viru-lence. In one case, flagella-specific phages were used togenerate phage-resistant non-motile mutants. Direct ino-culation of test plants showed that the non-motile mutantswere equally as virulent as the parent strain, and were stillable to attach to plant cells (Bradley et al., 1984). Anothergroup reported that Tn5 mutants deficient in motility andchemotaxis to root exudates of pea remained fully virulentwhen the inoculum was directly applied (Hawes and Smith,1989). Mutant bacteria indirectly applied were almost asvirulent as the parent on plants grown in sand but wereavirulent on soil-grown plants. This observation suggestseither that soil impeded root-exudate signals from reach-ing the bacteria, or that the bacteria were unable to

Molecular Microbiology (1997) 23(3), 579–590

Q 1997 Blackwell Science Ltd

Received 1 July, 1996; revised 15 November, 1996; accepted 20November, 1996. Present addresses: †Department of MedicalPathology, School of Medicine, University of California, Davis, Cali-fornia 95616, USA. ‡Department of Plant Pathology, Faculty of Agri-culture, Cairo University, Giza, Egypt. *For correspondence. E-mailcikado@ucdavis.edu; Tel. (916) 752 0325; Fax (916) 752 5674.

m

effectively reach the roots owing to the natural microbialfiltration capacity of soil. Amino acids such as arginineand valine, and simple sugars such as sucrose, glucoseand fructose are good chemoattractants (Loake et al.,1988). In one case, plant phenolics such as acetosyrin-gone were claimed to be chemoattractants specific toTi-plasmid-containing strains (Ashby et al., 1988). Aminoacids and sugars are natural exudates of plant roots, espe-cially from young seedlings (Schroth and Snyder, 1961).

To gain some insight into the role of flagella in facilitatinghost recognition and infection, we have identified, isolatedand characterized three flagella (fla) genes of A. tume-faciens C58. Mutations in each of these genes result indelayed motility and alteration in swimming behaviour.The deletion of flaA, flaB, and flaC genes results in anon-flagellated and non-motile phenotype, termed hereinas the ‘bald’ strain. Nucleotide sequencing revealed thepresence of three open reading frames (ORFs) corre-sponding to flaA, flaB and flaC. Primer-extension studiesshowed that these ORFs bear individual transcriptionalstart sites. Tests performed with the bald strain harbour-ing the Ti plasmid showed that such organisms remainedvirulent. However, the level of virulence of the bald strainwas consistently less than the parent wild-type strain asreflected by the distinct differences in tumour size, tumourfresh weight, and tumour number.

Results

Purification and sequence analysis of flagellin releasedfrom cells

Flagellins are frequently observed by SDS–PAGE analy-sis of the culture supernatant of bacterial cells. With A.tumefaciens cultures, two prominent proteins of 32 and33 kDa are frequently observed in the culture supernatant,and can be sheared off the cells by mechanical aggitation,e.g. vortexing (Fig. 1). The 33 kDa protein band is moreintense than the 32 kDa protein band, suggesting thatthere may be another protein superimposed. To verifythat these proteins were indeed flagellin, we transferredthem from the polyacrylamide gel to a polyvinylidenedifluoride (PVDF) membrane, and their amino acidsequences starting at their N-terminus were determined.The N-terminal amino acid sequence was MASILTNNNA-MAAL for the 32 kDa protein, and MTSI(I/L)TN(V/T)AAM-SALQ for the 33 kDa protein. The sequencing of the latterprotein revealed variations in the fifth and eighth aminoacid residues, which suggests that two proteins may haveco-migrated in the polyacrylamide gel. The sequencesare very similar to the amino acid N-terminus sequenceof flagellin encoded by flaA and flaB genes of Rhizobiummeliloti (Pleier and Schmitt, 1989). These results thereforesuggest that the major proteins released from A. tume-faciens cells in liquid culture represent flagellar proteins.

Identification and cloning of the fla genes

To further verify that these proteins are flagellin, we iden-tified and isolated the genes that encode the flagellarstructural proteins by screening a genomic library bycolony hybridization using the flaB gene from R. melilotias the hybridization probe. A hybridization-positive clonecontaining a 12.8 kb EcoRI DNA fragment was isolated.This fragment was subsequently subcloned as a 6.5 kbHindIII DNA fragment in cloning vector pTZ18R (Table1). This subcloned DNA fragment was sequenced, andsequentially arranged ORFs were found (Fig. 2). The pre-dicted amino acid sequence of the N-terminal region of thefirst and second ORFs exactly matched the sequencesdetermined for the purified 32 and 33 kDa proteins foundin the culture supernatant. The nucleotide sequence iden-tities are as follows: flaA with flaB, 71.1%; flaA with flaC,59.9%; and flaB with flaC, 70.4%. The complete nucleo-tide sequence has been deposited in the EMBL NucleotideSequence Database under the accession number X96435.

Based on the nucleotide sequences, flaA, flaB, and flaCeach contain a coding capacity for predicted proteins of307, 321, and 314 amino acid residues, and Mr of 31 637,32 966, and 32 770, respectively (Fig. 3). The predictedamino acid sequences of FlaA and FlaB of A. tumefaciensand FlaA and FlaB of R. meliloti show an average of 66%identity (Fig. 3). Interestingly, the A. tumefaciens FlaA andFlaB proteins are 80 amino acid residues shorter thanFlaA and FlaB proteins of R. meliloti and lack histidineand cysteine residues. This difference is observed in theabsence of corresponding amino acid residues spanningresidues 186–246 and residues 278–298 in FlaA, andresidues 208–270 in FlaB (Fig. 3).

fla promoters

RNA purified from A. tumefaciens cultures grown underthe same conditions as used for flagella purification was

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Fig. 1. Flagellar proteins of A. tumefaciens NT1RE cellsfractionated by SDS–PAGE and stained with Coomassie brilliantblue. Lane 1, protein molecular mass markers in kDa; lane 2,released flagellin proteins of 32 (lower) and 33 (upper) kDa bands.Note, the upper band contains two proteins of similar size, i.e.32 966 and 32 770 Da.

580 O. Chesnokova et al.

hybridized with an excess of single-stranded syntheticoligonucleotides labelled with 32P at their 58-termini, inorder to determine the transcriptional start of flaA, flaB,and flaC. For this hybridization, the synthetic oligonucleo-tide primers A3, B9, and C6 (see the Experimental pro-cedures) were extended by reverse transcriptase toproduce cDNA complementary to the RNA template, andthen treated with S1 nuclease. Protected fragments wereexamined by autoradiography following electrophoresisof samples on an 8% polyacrylamide–7 M urea sequen-cing gel (Sambrook et al., 1989). As shown in Fig. 4, thetranscription start (the nucleotide is shown by the arrow)for each fla gene is identified. Putative ¹10 and ¹35 con-sensus sequences of flaA and flaB (shown in bold) for as28 promoter are located upstream of each transcriptionstart (Fig. 5). For flaC, the putative s28 consensus boxes(underlined, not bold) were not detected by primer-

extension analyses. However, these analyses showedthat an additional putative promoter upstream of the tran-scription start of flaC contains closely spaced boxes atpositions ¹11 and ¹21 (bold gg and gc in Fig. 5) thatappear akin to consensus boxes of a s54 promoter(Kustu et al., 1989; Helmann, 1991). The transcriptionstart of flaC is variable because we observed that thestarting nucleotide is one of the three thymidines (Fig. 5,underlined t residues). This arrangement is similar toclass III and class IV fla gene promoters of Caulobactercrescentus (Wu et al., 1995). In our case, the conservedGC is 11 bp upstream of the transcriptional start and theconserved GG doublet is 10 bp farther upstream; suchmotifs are considered to be the minimal requirement fors54 promoters (Kustu et al., 1989).

Situated upstream of these putative promoters areinverted repeats (Fig. 5), suggesting potential hairpin-loopstructures that may play a role in fla gene expressione.g. possibly as transcriptional terminators. Another invertedrepeat exists 7 nucleotides downstream of flaC.

Generation of flagella-defective mutants

Mutant strains defective in flagella organization or motilitywere constructed by marker-integration mutagenesis(Kamoun et al., 1992a). Using the sequence informationshown in Fig. 3, and oligonucleotide primers respectiveto each fla gene sequence, a DNA fragment internal toeach fla gene was cloned in pUCD5701 and transformedinto E. coli S17-1(lpir ). Plasmid DNA purified from thetransformant was transferred into A. tumefaciens NT1REby electroporation and screening was performed to selectfor kanamycin-resistant transformants: only those har-bouring an integrated plasmid would be resistant to kana-mycin. Each fla gene mutant was tested for motility andeach mutant was defective in their ability to rapidly swimout from the original site of inoculation in swarm-agarplates. However, after 24–48 h incubation at 288C allthree fla mutants showed bacterial surface translocationknown as twitching motility observed in other organisms

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Table 1. Bacterial strains and plasmids used in this study.

Strain/Plasmid Description Source/Reference

Strain

A. tumefaciensNT1RE RmR EmR pTi¹ motþ Kao et al. (1982)NT1REB RmR EmR pTi¹ mot¹

DflaABC (bald strain)This work

E. coliXL1-Blue recA1 endA1 gyrA96 thi-1

hsdR17 supE44 relA1 lac(F8::Tn10 proAB lacI q

lacZDM15 )

Stratagene

S17-1(lpir) lacI lacZ Simon et al.(1983)

Plasmid

pTZ18R ApR U.S. BiochemicalspBluescript-II

SK¹

ApR lacZ promoter Stratagene

pSa151 KmR SpR pSa ori Tait et al. (1983)pJK270 pTiC58trac::Tn5(KmR) Kao et al. (1982)pJQ200uc1 GmR sacB Quandt and

Haynes (1993)pUCD5701 ApR R6K ori pir¹ Cha et al. (1996)pRU929 TcR, flaB from R. meliloti Pleier and Schmitt

(1991)pUCD5300 12.8 kb EcoRI-generated

flaABC-containing DNAfragment cloned inpBluescript-II SK¹

This work

pUCD5302 6.5 kb HindIII-generatedflaABC-containing DNAfragment cloned in pTZ18R

This work

pUCD5370 6.5 kb HindIII-generatedflaABC-containing DNAfragment cloned inpJQ200uc1

This work

pUCD5371 flaABC-deletion derivative ofpUCD5370

This work

RmR, rifampicin resistant; EmR, erythromycin resistant; ApR, ampicil-lin resistant; KmR, kanamycin resistant; SpR, spectnomycin resistant;GmR, gentamicin resistant; TcR, tetracycline resistant.

Fig. 2. Physical map of flaA, flaB, and flaC indicated as boldarrows on a 6.5 kb DNA fragment. The bold arrows indicate thedirection of transcription. The line beneath the fla genes is thesection deleted in the A. tumefaciens NT1RE genome to generatethe bald strain NT1REB. The restriction-endonuclease sites areabbreviated as follows: H, HindIII; P, Pst I; A, AseI; B, BamHI;D, DraI; C, ClaI; K, KpnI; N, Not I S, SacI; Sp, SspI.

Agrobacterium fla genes affect virulence 581

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Fig

.3.

Com

pari

sons

betw

een

the

pred

icte

dam

ino

acid

sequ

ence

sfo

rF

laA

,F

laB

,an

dF

laC

ofA

.tu

mef

acie

ns(A

t)an

dkn

own

Fla

Aan

dF

laB

amin

oac

idse

quen

ces

ofR

.m

elilo

ti(R

m)

(Ple

ier

and

Sch

mitt

,19

89).

Seq

uenc

eco

mpa

riso

nw

aspe

rfor

med

byus

ing

the

BE

ST

FIT

Pro

gram

(GC

Gpa

ckag

e,W

isco

nsin

).T

heve

rtic

allin

esin

dica

teid

entic

alam

ino

acid

resi

dues

,w

here

asth

eco

lons

(:)

and

peri

ods

(.)

are

com

pari

son

valu

esof

0.50

or0.

10,

resp

ectiv

ely,

acco

rdin

gto

the

prog

ram

.

582 O. Chesnokova et al.

(Whitchurch and Mattick, 1994) (Fig. 6). The swimmingbehaviour of these mutants was also examined by phase-contrast light microscopy. Compared with parental wild-type cells, flaB and flaC mutants exhibited defectiveswimming behaviour by tumbling incessantly. Although,no translational motility associated with the swimmingbehaviour of the wild-type strain was observed with anyof the three mutants, delayed motility was observed.

This type of flagella-independent motility is associatedwith type 4 pili (Alm et al., 1996), therefore, part of theswimming behaviour might be accounted for by pili-mediated twitching motility. Electron microscopic exami-nation of these mutants revealed the following pheno-types: flaA mutants mainly produced no flagella (withvestigial flagella stubs infrequently observed), whereasflaB and flaC mutants elaborated defective flagella,

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Fig. 4. Identification of the transcription-initiation site by primer-extension analysis. The initiating nucleotide for each transcript was mappedusing oligonucleotide primers complementary to sequences within and extending out from the amino-terminal region of each fla ORF (see theExperimental procedures). Primer A3 was used for flaA, B9 for flaB, and C6 for flaC. Each primer was end-labelled and annealed to RNAfrom wild-type A. tumefaciens. Reverse transcriptase was then added and the products analysed on an 8% polyacrylamide–7 M ureasequence gel. The transcriptional start site for each fla gene is indicated by the arrow next to the sequencing ladder.

Fig. 5. Analysis of the regulatory sequencesupstream of the transcription start in thenon-coding region of each fla gene. Thepotential s28 promoters are indicated byunderlined, bold, lower-case letters groupedupstream of each fla gene, and the putatives54 promoter consisting of the closely spacedGG and GC in bold lower-case letters isupstream of flaC. The transcriptional startnucleotide (determined by primer-extensionanalysis; see Fig. 4) upstream of each flagene is in bold and underlined. Thetranscription start of flaC shows a thymidinetriplet instead of a single nucleotide becauseof primer-extension ambiguity. Invertedrepeated sequences are indicated with linedarrows. The ribosome-binding site isdesignated RBS.

Agrobacterium fla genes affect virulence 583

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Fig. 6. Motility assay of the fla and baldmutant strains. Close-up view of parentalwild-type strain (wt), mutant strains flaA (A),flaB (B), flaC (C), and the bald strain NT1REB(D) plated on 0.4% swarm agar, andincubated for 48 h at 258C.

Fig. 7. a. Electron micrograph of negatively stained flagellamutants: A, flaA; B, flaB; and C, flaC; and parental strain, P. Barsshown in panels P, A, and C equal 10 mm. The bar shown in panelB equals 1 mm.b. Enlarged view of parental strain (A) and bald mutant strainNT1REB (B).

584 O. Chesnokova et al.

bearing kinks and sharp right-angle bends not seen withwild-type flagella (Fig. 7). The latter two strains producedpili, while the flaA strain did not.

Because flaB and flaC mutants still produced flagella,which were albeit defective in structural integrity, weexploited the sequentially arranged fla genes in A. tume-faciens to construct a mutant strain containing a deletionof all three fla genes (Fig. 2). To facilitate this construc-tion, we subcloned the 6.5 kb HindIII DNA fragmentcontaining the three fla genes into the cloning vectorpJQ200uc1. A 3.9 kb Pst I fragment containing all of flaA,flaB, and 95.8% of flaC was then removed, and theremaining DNA was religated and transformed into E.coli S17-1(lpir ). The resulting construct was transferredinto A. tumefaciens NT1RE by electroporation and platedonto induction medium containing rifampicin and gentimi-cin. Gentimicin-resistant transformants were then platedon medium 523 (see the Experimental procedures) con-taining rifampicin. Because a single cross-over event willstill contain the sacB gene of the vector plasmid, sucrose,which is a regular component of medium 523, would belethal to those cells. On the other hand, recombinantsthat had undergone double cross-over events would havelost sacB and therefore would be able to grow on mediumcontaining sucrose (Kamoun et al., 1992b). Thus, we wereable to obtain a series of mutant strains that were unableto produce flagella and were non-motile. One of thesestrains was selected at random and is referred to as the‘bald’ strain, NT1REB. The absence of the fla genes wasverified (i) by Southern blot hybridization using the deletedfla gene fragment as the probe, (ii) by analysing the culturesupernatant for the Fla proteins, i.e. the loss of each Flaprotein occurs for each corresponding fla mutant (PAGEdata not shown), and (iii) by electron microscopy (Fig. 7).In addition, motility assays of the bald strain verified thatit is stably non-motile in swarm-agar plates (Fig. 6). Whenthis strain was complemented with the wild-type parentalfla genes, the strain regained full motility and producedflagella resembling those of the wild-type strain (data notshown).

Effect of bald strain on virulence

Despite the earlier observations that non-motile mutantsremain virulent when inoculated directly onto test plants(Bradley et al., 1984; Hawes and Smith, 1989), no quanti-tative assays were performed to determine whether subtledifferences in virulence might be present between flagella-bearing motile strains and flagella-free non-motile strainsof A. tumefaciens. When we tested the tumour-inducingability of the bald strain NT1REB on sunflower plants, atest host strain that is useful for virulence assessmentsbased on tumour size and tumour fresh weight (Langleyand Kado, 1972), and on red-potato tuber disks for

quantitative estimates of virulence (Pueppke and Benny,1981), we found that the size and weight of the tumourswere consistently smaller than that induced by the wild-type C58 strain (Fig. 8). We found an average reductionin both the size and fresh weight of 36% even when weused Jimson weed (Datura stramonium) as a test host(Fig. 8). Of five replicated trials, the average tumour freshweight was 0.39 6 0.12 g for NT1RE(pJK270), and 0.25 6

0.11 g for NT1REB(pJK270). The difference observedwas consistent in five independent trials. This attenuationof virulence was confirmed by the quantitative assay whichshowed an average reduction in the tumour numbers of38% (Table 2).

For motility and swarm behaviour, the bald strain andparental wild-type strain were reisolated from 10-week-old crown-gall tumours on experimentally inoculated Jim-son weed, and tested for motility and swarming onswarm-agar plates. No changes in motility and swarmingwere observed; the reisolated bald strain remained non-motile, while the wild-type strain was fully motile andswarmed (data not shown). Thus, the bald strain remainsstably non-motile, and does not undergo phase transitioninto a motile form when it comes in contact with hostcells. Compared with the bald strain, only the flaA mutantshowed slight decrease in tumorigenicity, while the baldstrain showed an appreciable decrease in virulence(Table 2).

Discussion

The thrust of our research is directed towards developing aflagella-free strain of A. tumefaciens in order to facilitate

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Fig. 8. Crown-gall tumours produced on sunflower and jimsonweed by wild-type NT1RE (pJK270), and bald strainNT1REB(pJK270).A. Sunflower stem sections inoculated with: 1, water;2, NT1RE(pJK270); and 3, NT1REB(pJK270).B. Jimson-weed stem sections inoculated with: 1, water; 2,NT1RE(pJK270); and 3, NT1REB(pJK270).

Agrobacterium fla genes affect virulence 585

the visual observation by high-resolution microscopy ofpotential pili involved in conjugative transfer of the T-DNAto plant cells. The initial objective was therefore directedtoward identifying and isolating flagella genes so thatthey could be removed from the bacterial chromosomeby deletion mutations. Flagellar proteins are frequentlythe major proteins that occur in culture supernatantsbecause flagellated bacteria shed their flagella throughmechanical agitation. Our analysis of the prominent pro-teins appearing in the culture supernatant proved themto be those from flagella because of the following: (i) theamino acid sequence of the N-terminus of these proteinsshowed homologies to the published FlaA and FlaBsequence of R. meliloti (Pleier and Schmitt, 1989); (ii)the nucleotide sequence of the isolated ORFs corre-sponding to flaA and flaB genes were similar (by anaverage of 66% identity) to those of R. meliloti ; and (iii)the N-terminal amino acid sequence predicted from theseORFs matched exactly with the amino acid sequenceanalysis of the two isolated proteins. Sequence analysisof the cloned DNA bearing flaA and flaB also revealed athird ORF whose sequence was similar to flaA and flaBbut contained dissimilar sequences in the central regionof the ORF. Genetic analysis of this ORF indicated thata third flagella gene is present. Marker-integration muta-tions resulted in either the loss of motility (flaA), or inseverely altered swimming behaviour, i.e. tumbling (flaBand flaC ), as observed by light microscopy (data notshown). The latter phenotype may either be caused bythe abnormably bent or kinked flagella as observed byelectron microscopy (Fig. 7), or to an unknown defect.For example, Che¹ mutant strains of Salmonella typhi-murium showed a tumbly swimming behaviour (Jonesand Falkow, 1994). Whether or not flaB and flaC mutantstrains of Agrobacterium are affected in a che locusremains to be determined. Deletion of the fla genes

resulted in a non-motile, bald strain (NT1REB), and elec-tron microscopic examination of this strain confirmed theabsence of flagella.

Some insight into gene regulation was gained byexamining the potential promoter of each fla gene relativeto its transcriptional start site. These studies revealed aconcensus sequence and spacing arrangement for a puta-tive s28 promoter (Fig. 3). Sigma 28 is known to promotetranscription of genes whose products are required formotility and chemotaxis among diverse bacteria (Hel-mann, 1991). The sequence upstream of flaC containsan additional concensus box and spacing arrangementfor a s54 promoter, suggesting that flaC expression mightbe under dual regulatory control by means of two pro-moters whose s factors may be responsive to specificenvironmental or physiological conditions. Such is thecase for the nfeC gene that is involved in conferringnodulation efficiency and competitiveness on Bradyrhizo-bium japonicum (Chun and Stacey, 1994). Two closelyspaced and independently regulated promoters are locatedupstream of nfeC in this organism. In our study, it wasinteresting to find the potential for two promoters becauses54 is known to control flagella biosynthesis in a variety ofbacteria (Kustu et al., 1989). Clearly, further studies usingreporter genes will help elucidate these preliminary find-ings that resulted from the sequence characterizationreported in this article.

Motility and chemotaxis are generally attributed toflagella and their motors; less recognized is their role invirulence. Motility-defective, aflagellated bacteria havebeen reported to be affected in virulence because ofabberations in their host adherence, invasion mechanism,or unknown factors. These include aflagellated Vibriocholerae (Attridge and Rowley, 1983), Campylobacter jejuni(Wassenaar et al., 1991), Borrelia burgdorferi (Sadzieneet al., 1991), Clostridium chauvoei (Tamura et al., 1995),

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Table 2. Quantitative tumour assay forvirulence of bald strain NT1REB. Test

groupStrain withpJK270

No. ofdisks

No. oftumours

Tumours per disk(6SD)a

Relativevirulence (%)

1 NT1RE 13 180 13.8 6 4.9 100NT1REB 15 138 9.2 6 4.0 66.5

2 NT1RE 20 417 20.9 6 9.0 100NT1REB 20 234 11.7 6 6.2 56.1

3 NT1RE 33 519 15.7 6 8.5 100NT1REB 36 396 11.0 6 7.6 70.0

4 NT1RE 25 611 24.4 6 12.8 100NT1REB 29 521 17.8 6 7.1 73.5

5 NT1RE 32 985 30.8 6 14.7 100NT1REB 33 757 23.6 6 12.0 75.8

a. Student’s t-test was used to determine the probability (P ) that the difference between thepaired mean numbers are significant. A confidence limit of P < 0.05 was found for each pairedsample.

586 O. Chesnokova et al.

S. typhimurium (Jones et al., 1992; Khoramianfalsafi et al.,1990), Helicobacter pylori (Eaton et al., 1992), and Bacil-lus thuringiensis (Zhang et al., 1993). With regard toplant pathogens, correlations between chemotaxis, motil-ity, and virulence were found for Xanthomonas campestrispv. campestris (Kamoun and Kado, 1990), Pseudomonassyringae pv. phaseolicola (Panopoulos and Schroth, 1974),and Erwinia carotovora ssp. atroseptica (Mulholland et al.,1993).

Studies on aflagellated A. tumefaciens have not beenreported, although earlier brief reports on non-motilemutants obtained by screening for resistance to flagella-specific phage suggested that there were no differencesbetween the parent and non-motile derivative in the abilityto bind suspension cells and cause tumours (Bradley et al.,1984).

On the other hand, by using quantitative analyses, wefound statistically significant differences in virulencebetween our bald strain and its parental strain (Table 2).Another laboratory (Hawes and Smith, 1989) has shownthat virulence was drastically affected by soil type whena suspension of non-motile A. tumefaciens cells wasadded distal to the seedlings. When test plants weregrown in sand, no difference in virulence was observed,but in clay soil the same mutant strains failed to causetumours, presumably because of their inability to effec-tively reach the host plant (Hawes and Smith, 1989).Shaw et al. (1991) reported chemotaxis by A. tumefacienstowards plant phenolics such as acetosyringone, andobserved that non-motile mutants were non-chemotacticand incapable of colonizing the roots of young potatoseedlings. These observations would suggest that motility,and presumably flagella, may be a necessary componentfor ecological fitness in natural environments.

Besides their role in motility, it might be possible thatflagella are involved in stability maintenance (e.g., stiltingin contrast to direct binding) during host–pathogen inter-action, and/or enhancing swarming at the site of primaryinfection. Although we have not provided evidence hereto support these notions, the circumthecal arrangementof the flagella near one end of the A. tumefaciens cellcould play a structural role in facilitating the intercellularcomplex formation, in addition to directional swimming(e.g. towards the roots). Electron microscopic examinationof plant cells interacting with A. tumefaciens has revealedthe bacterial cells attached in a polar fashion (Smith andHindley, 1978; Matthysee et al., 1981; M. C. Hawes,unpublished). Certain octopine strains such as A6 showboth polar and lateral attachments (Matthysse et al.,1981). The loss of flagella might therefore cause a decreasein stable attachment maintenance and, thus, a concomi-tant decrease in virulence. The stabilization of matingpairs in F-plasmid-mediated conjugation is essential forDNA transfer in Escherichia coli (Frost et al., 1994).

Hence, the observed polar attachment of A. tumefaciensto plant cells might be a necessary prelude leadingtowards T-DNA transfer from one end of the bacilliformAgrobacterium cell, a process which is believed to be bymeans of a conjugative mechanism (reviewed recently inKado, 1993; Lessl and Lanka, 1994). Our early studies(Kado, 1976) and those by Tanaka (1985) observed longfilamentous structures reminiscent of pili that were alwayson one end of the cell. A growing body of evidence sug-gests that the mechanism of T-DNA transfer from Agro-bacterium to plant cells is via a sex-pilus-like structure(Shirasu and Kado, 1993; Kado, 1994; Fullner et al.,1996). The polar means of DNA transfer stabilized byflagella therefore might be a logical way for Agrobacteriumto deliver its genetic material to host plants.

The alternative explanation for reduced virulence by thebald strain would be the loss of swarming behaviour, whichwould reduce the size of the inoculum and therefore con-tribute directly to the lowering of virulence because ofinoculum insufficiency.

Whatever is the case, the reduction in virulence appearsto be the result of a dysfunctional flagella system.

Experimental procedures

Bacterial strains, plasmids, and culture conditions

Strains and plasmids are described in Table 1. A. tumefaciensNT1RE that is free of the Ti plasmid and resistant to100 mg ml¹1 rifampicin (Rm) and 150 mg ml¹1 erythromycin(Em) was constructed by Kao et al. (1982). E. coli XL-1 blue(Stratagene Cloning Systems), and strain S17-1(lpir ) (Simonet al., 1983) were used as recipients of the following plasmids:pTZ18R (U.S. Biochemicals) was used for cloning; pRU929containing flaB of R. meliloti (Pleier and Schmitt, 1991), kindlyprovided by Rudiger Schmitt, was used as the source of theflaB gene that was used as a DNA probe; pSa151 (Tait etal., 1983), containing NT1RE flaA, flaB, and flaC genes, wasused in complementation tests; and pJQ200uc1, which wasprovided by Michael F. Haynes (Quandt and Haynes, 1993),was modified into pUCD5701 (Cha et al., 1996) containingthe origin of DNA replication of R6K, the levan sucrase genesacB, and the aminoglycoside phosphotransferase aph geneto generate deletions, according to the method of Kamounet al. (1992b).

E. coli strains were grown in Luria–Bertani (LB) broth at378C with shaking (250 r.p.m.). A. tumefaciens strains weregrown at 308C with shaking (250 r.p.m.) in medium 523(10 g l¹1 sucrose, 8 g l ¹1 tryptone (Difco), 4 g l ¹1 yeast extract,3 g l¹1 K2HPO4, 0.3 g l¹1 MgSO47H2O) or in induction medium(Rogowsky et al., 1987). Where indicated, gentimicin wasused at 20 mg ml¹1.

DNA preparation

Chromosomal DNA was prepared from exponentially growingA. tumefaciens NT1RE cells and used to construct a genomiclibrary (Cooley et al., 1991). Plasmid DNA was prepared bythe method of Kado and Liu (1981), except that the last

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Agrobacterium fla genes affect virulence 587

DNA precipitation step included the addition of 400 ml of 3Msodium acetate to 100 ml of the DNA preparation in the solu-tion taken from the partitioned aqueous phase above thechloroform–phenol layer, followed by precipitation of the plas-mid DNA with two volumes of ice-cold ethanol. The precipitatewas washed at least four times with 70% ethanol to removeresidual detergent. Purified chromosomal DNA was digestedwith restriction endonucleases, and the resulting DNA frag-ments were separated by electrophoresis in a 0.8% agarosegel in E buffer (40 mM Tris-acetate, 4 mM disodium EDTA,pH 7.9) and stained with 0.5 mg ml¹1 of ethidium bromide.Fractionated DNA bands visualized by long-wave UV lightwere excised with a scalpel from the agarose gel and purifiedusing Geneclean II according to the procedure described bythe manufacturer (Bio 101).

Cloning, sequencing, and determation of thetranscriptional start nucleotide of flagellin genes

Chromosomal DNA was cleaved with EcoRI according to themanufacturer’s instructions (Boehringer Mannheim) and frac-tionated by electrophoresis in a 1% agarose gel as describedabove. An EcoRI-generated 12.8 kb DNA fragment containingthe fla genes was cloned in pTZ18R, and further subclonedinto pTZ18R as a 6.5 kb HindIII insert. Nucleotide sequencesof overlapping clones of this insert were determined by themethod of Chen and Seeburg (1985). Nucleotide sequencesin both directions of overlapping deletion clones were deter-mined by the procedure of Sanger et al. (1977), using 7-deaza dGTP to clarify regions previously obscured by GCcompression (Mizusawa et al., 1986). Sequence data werecompiled and comparative analyses were made using theGCG program (University of Wisconsin, Madison), the BLASTN,BLASTX, and BESTFIT programs developed by the NationalCentre for Biotechnology Information (Altschul et al.,1990; Devereux et al., 1984; Gish and States, 1993), andthe GENECOMPAR program (Applied Maths Department, Kor-trijk, Belgium).

The transcriptional start sites of each fla gene was deter-mined by primer-extension analysis (Sambrook et al., 1989),utilizing the following synthetic oligonucleotides as primers:A3 (58-CTTGCCATAATAAATGTG-38: position 308–291)for flaA; and B9 (58-TGCTCGTCATAGTAGTGT-38: position1637–1620) for flaB; and C6 (58-AATACTTGTCATAATT-GC-38: position 2843–2826) for flaC. These primers werelabelled with [g-32P]-dATP catalysed by T4 kinase (Sam-brook et al., 1989). RNA was purified from A. tumefacienscells grown under the same conditions as those for flagellardevelopment and isolation. Primer–RNA hybridizations wereperformed at 58C below the calculated Tm of each primer.Control hybridization reactions contained no RNA. Synthesisof cDNA employed AMV reverse transcriptase (Promega) andthe reaction was stopped with S1 nuclease treatment for 1.5 hat 138C. The product of the reaction was extracted with25:24:1 phenol:chloroform:isoamylalcohol (v/v/v), and thenucleic acid in the upper aqueous layer was precipitatedwith 95% ethanol, washed with 70% ethanol, dried, and resus-pended in formamide loading buffer (80% formamide, 10 mMEDTA, pH 8.0) (Sambrook et al., 1989). Aliquots (5 ml) weresubjected to electrophoresis on an 8% polyacrylamide sequen-cing gel containing 7 M urea.

Flagellin purification and sequencing

Flagellin was purified from NT1RE cells grown on medium523 agar plates at 308C for 3 d. Bacterial cells were resus-pended in 5 ml distilled water and vortexed for 3 × 15 sec, todetach flagella. Cells were removed by centrifugation at5800 ×g for 10 min at 48C. The supernatant was recentrifugedat 9600 ×g for 15 min at 48C, and the resulting supernatantwas centrifuged at 50 000 ×g for 40 min at 48C. The pelletedpreparation was composed of 95% flagella, as determinedby solubilizing the flagellin in 1% SDS at 1008C for 5 min,and fractionating the polypeptide electrophoretically in a7.5% SDS–PAGE gel. The flagellin-bearing band was trans-ferred by electroblotting onto a polyvinylidene fluoride mem-brane (PVDF-PLUS, Micro Separations, Inc.). The segmentcontaining the flagellin was placed into a Blott cartridge(Perkin–Elmer), and the N-terminal amino acid sequencewas determined directly from the membrane by automatedEdman degradation (sequential phenyl isothiocyanatedegradation) in an Applied Biosystems model 470A gas-phasesequencer equipped with an on-line high-performance liquidchromatography (HPLC) system and corresponding ABI soft-ware (Perkin–Elmer). Amino acids were identified against aprofile of amino acid standards.

Construction of marker-integration mutants

Internal sequences of the flaA, flaB, and flaC genes werecloned in pUCD5701 and transformed into E. coli S17-1(lpir ). Plasmid DNA from transformants harbouring oneof each of these fla genes was purified as described above,and transformed into NT1RE by electroporation (Cooley etal., 1991). Because pUCD5701 contains the origin of DNAreplication of plasmid R6K that requires the 35 kDa replicationprotein p encoded by the pir gene usually situated next to theorigin (Filutowicz et al., 1986), it is unable to replicate inNT1RE in the absence of the pir gene, and only plasmidsthat integrate into the chromosome by a single cross-overevent within the internal fla gene fragment will confer theresistance to kanamycin (50 mg ml¹1) by the aph gene inpUCD5701. Each marker integration was verified by restric-tion-fragment mapping and blot hybridization using pUCD5701as the probe.

Electron microscopy

Flagella-bearing NT1RE cells grown on medium 523 agarwere sampled by placing a droplet of 25 ml of distilled wateronto the edges of single colonies, incubated for 30 min in amoist chamber at 238C, and retrieved by touching the dropletwith a carbon-coated 300 mesh copper grid (Ernest F.Fullam, Inc.). The sample on the grid was air dried and nega-tively stained with 20 ml of 3% uranyl acetate, pH 7.0. The gridwas floated on ammonium acetate buffer, pH 7.0, for 1 h, andair dried. Each preparation was viewed in a Phillips EM410electron microscope at 80 kV.

Virulence assay

Virulence of NT1RE and the respective fla isogenic mutants,

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

588 O. Chesnokova et al.

each containing pTiC58 (as pJK270 which contains a Tn5insertional marker in a region near the left border of theT-DNA) (Kao et al., 1982), was tested on sunflower (Heli-anthus annuus cv. Russian Mammouth) and on Jimsonweed (D. stramonium) as described previously (Langley andKado, 1972). The stems of five-week-old seedlings of theseplants were inoculated with 107 cells ml¹1 of test and controlinocula, and four-week-old crown-gall tumours appearing atthe inoculation site were cut away from the stem with a scapel.The fresh weights of the tumour tissue were measured on ananalytical balance (Sartorius). Quantitative tumour assayswere performed using disks (10 mm diameter × 4 mm thick)obtained from cored red potato tubers (Solanum tuberosum),which were inoculated with 107 cells ml¹1 and placed on 1.5%water agar according to the method of Pueppke and Benny(1981). Tumours were enumerated after 2–3 weeks incuba-tion of the inoculated disks at room temperature in the dark.

Acknowledgements

This research was supported by Public Health Service GrantGM45550 from the National Institutes of Health. M.S.M. wassupported by a senior fellowship from the American Middle-East Peace Fellowship Program.

References

Alm, R.A., Hallinan, J.P., Watson, A.A., and Mattick, J.S.(1996) Fimbrial biogenesis genes of Pseudomonas aerugi-nosa: pilW and pilX increase the similarity of type 4 fim-briae to the GSP protein-secretion system and pilY1encodes gonococcal PilC homologue. Mol Microbiol 22:161–173.

Altschul, S., Gish, W., Miller, W., Myers, E., and Lipman, D.(1990) Basic local alignment search tool. J Mol Biol 215:403–410.

Ashby, A.M., Watson, M.D., Loake, G.J., and Shaw, C.H.(1988) Ti plasmid-specified chemotaxis of Agrobacteriumtumefaciens C58C toward vir -inducing phenolic com-pounds and soluble factors from monocotyledonous anddicotyledonous plants. J Bacteriol 170: 4181–4187.

Attridge, S.R., and Rowley, D. (1983) The role of the flagel-lum in the adherence of Vibrio cholerae. J Infect Dis 147:865–872.

Bradley, D.E., Douglas, C.J., and Peschon, J. (1984) Flagella-specific bacteriophages of Agrobacterium tumefaciens:demonstration of virulence of non-motile mutants. Can JMicrobiol 30: 676–681.

Cha, J.-S., Pujol, C., and Kado, C.I. (1996) Identification andcharacterization of a Pantoea citrea gene encoding glu-cose dehydrogenase that is involved in causing the pinkdisease of pineapple. Appl Environ Microbiol 63: in press.

Chen, E.Y., and Seeburg, P.H. (1985) Supercoil sequencing:a fast and simple method for sequencing plasmid DNA.DNA 4: 165–170.

Chun, J.-Y., and Stacey, G. (1994) A Bradyrhizobium japoni-cum gene essential for nodulation competitiveness isdifferentially regulated from two promoters. Mol Plant–Microbe Interact 7: 248–255.

Cooley, M.B., D’Souza, M.R., and Kado, C.I. (1991) The virCand virD operons of the Agrobacterium Ti plasmid are

regulated by the ros chromosomal gene: analysis of thecloned ros gene. J Bacteriol 173: 2608–2616.

Devereux, J., Haeberli, P., and Smithies, O. (1984) A com-prehensive set of sequence analysis programs for theVAX. Nucl Acids Res 12: 387–395.

Eaton, K.A., Morgan, D.R., and Krakowka, S. (1992) Motilityas a factor in the colonisation of gnotobiotic piglets by Heli-cobacter pylori. J Med Microbiol 37: 123–127.

Filutowicz, M., McEachern, M.J., and Helinski, D.R. (1986)Positive and negative roles of an initiator protein at an ori-gin of replication. Proc Natl Acad Sci USA 83: 9645–9649.

Frost, L.S., Ippen-Ihler, K., and Skurray, R.A. (1994) Analysisof the sequence and gene products of the transfer region ofthe F sex factor. Microbiol Rev 58: 162–210.

Fullner, K.J., Lara, J.C., and Nester, E.W. (1996) Pilusassembly by Agrobacterium T-DNA transfer genes. Science273: 1107–1109.

Gish, W., and States, D. (1993) Identification of protein cod-ing regions by database similarity search. Nature Genet 3:266–272.

Hawes, M.C., and Smith, L.Y. (1989) Requirement for chemo-taxis in pathogenicity of Agrobacterium tumefaciens on rootsof soil-grown pea plants. J Bacteriol 171: 5668–5671.

Helmann, J.D. (1991) Alternative sigma factors and the regu-lation of flagellar gene expression. Mol Microbiol 5: 2875–2882.

Jones, B.D., and Falkow, S. (1994) Phenotypic and geneticaspects of host cell invsion by Salmonella species. In Mol-ecular Mechanisms of Bacterial Virulence. Kado, C.I., andCrosa, J.H. (eds). Dordrecht: Kluwer Academic Publishers,pp. 3–16.

Jones, B.D., Lee, C.A., and Falkow, S. (1992) Invasion ofSalmonella typhimurium is affected by the direction of fla-gellar rotation. Infect Immun 60: 2475–2480.

Kado, C.I. (1976) The tumor-inducing substance of Agro-bacterium tumefaciens. Annu Rev Phytopathol 14: 265–308.

Kado, C.I. (1992) Plant pathogenic bacteria. In The Prokary-otes. 2nd edn. Vol. 1. Balows, I.A., Truper, H.G., Dworkin,M., Harder, W., and Schleifer, K.-H. (eds). New York:Springer–Verlag, pp. 659–674.

Kado, C.I. (1993) Agrobacterium-mediated transfer and stableincorporation of foreign genes in plants. In Bacterial Conju-gation. Clewell, D.B. (ed.). New York: Plenum Press, pp.243–254.

Kado, C.I. (1994) T-DNA transfer to plants is mediated bypilus-like apparatus encoded by the Ti plasmid virBoperon. Adv Plant Biotech 4: 23–26.

Kado, C.I., and Liu, S.-T. (1981) Rapid procedure for detec-tion and isolation of large and small plasmids. J Bacteriol145: 1365–1373.

Kamoun, S., and Kado, C.I. (1990) Phenotypic switchingaffecting chemotaxis, xanthan production, and virulencein Xanthomonas campestris. Appl Environ Microbiol 56:3855–3860.

Kamoun, S., Kamdar, H.V., Tola, E., and Kado, C.I. (1992a)Incompatible interactions between crucifers and Xanthomo-nas campestris involve a vascular hypersensitive response:role of the hrpX locus. Mol Plant–Microbe Interact 5: 22–33.

Kamoun, S., Tola, E., Kamdar, H.V., and Kado, C.I. (1992b)Rapid generation of directed and unmarked deletions inXanthomonas. Mol Microbiol 6: 809–816.

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Agrobacterium fla genes affect virulence 589

Q 1997 Blackwell Science Ltd, Molecular Microbiology, 23, 579–590

Kao, J.C., Perry, K.L., and Kado, C.I. (1982) Indoleaceticacid complementation and its relation to host range speci-fying genes on the Ti plasmid of Agrobacterium tume-faciens. Mol Gen Genet 188: 425–432.

Khoramianfalsafi, T., Harayama, S., Kutsukake, K., andPechere, J.C. (1990) Effect of motility and chemotaxis onthe invasion of Salmonella typhimurium into HeLa cells.Microb Pathog 9: 47–53.

Kustu, S., Santero, E., Keener, J., Popham, D., and Weiss,D. (1989) Expression of s54 (ntrA)-dependent genes isprobably united by a common mechanism. Microbiol Rev53: 367–376.

Langley, R.A., and Kado, C.I. (1972) Studies on Agrobac-terium tumefaciens, conditions for mutagenesis by N-methyl-N 8-nitro-N-nitrosoguanidine and relationships ofA. tumefaciens mutants to crown-gall tumor induction.Mutat Res 14: 277–286.

Lessl, M., and Lanka, E. (1994) Common mechanisms inbacterial conjugation and Ti-mediated T-DNA transfer toplant cells. Cell 7: 321–324.

Loake, G.J., Ashby, A.M., and Shaw, C.H. (1988) Attractionof Agrobacterium tumefaciens C58C towards sugars involvesa highly sensitive chemotaxis system. J Gen Microbiol 134:1427–1432.

Matthysse, A.G., Holmes, K.V., and Gurlitz, R.H.G. (1981)Elaboration of cellulose fibrils by Agrobacterium tume-faciens during attachment to carrot cells. J Bacteriol 145:583–595.

Mizusawa, S., Nishimura, S., and Seela, F. (1986) Improve-ment of the dideoxy chain termination method of DNAsequencing by use of deoxy-7-deazaguanosine triphos-phate in place of dGTP. Nucl Acids Res 14: 1319–1324.

Mulholland, V., Hinton, J.C.D. Sidebotham, J., Toth, I.K.,Hyman, L.J., Perombelon, M.C.M., Reeves, P.J., and Sal-mond, G.P.C. (1993) A pleiotropic reduced virulence(Rvir ¹) mutant of Erwinia carotovora subspecies atro-septica is defective in flagella assembly proteins that areconserved in plant and animal bacterial pathogens. MolMicrobiol 9: 343–356.

Panopoulos, N., and Schroth, M.N. (1974) Role of flagellamotility in the invasion of bean leaves by Pseudomonasphaseolicola. Phytopathology 64: 1389–1397.

Pleier, E., and Schmitt, R. (1989) Identification and sequenceanalysis of two related flagellin genes in Rhizobium meliloti.J Bacteriol 171: 1467–1475.

Pleier, E., and Schmitt, R. (1991) Expression of two Rhizo-bium meliloti flagellin genes and their contribution to thecomplex filament structure. J Bacteriol 173: 2077–2085.

Pueppke, S.G., and Benny, U.K. (1981) Induction of tumorson Solanum tuberosum L. by Agrobacterium: quantitativeanalysis, inhibition by carbohydrates, and virulence ofselected strains. Physiol Plant Pathol 18: 169–179.

Quandt, J., and Haynes, M.F. (1993) Versatile suicide vec-tors which allow direct selection for gene replacement inGram-negative bacteria. Gene 127: 15–21.

Rogowsky, P.M., Close, T.J., Chimera, J.A., Shaw, J.J., andKado, C.I. (1987) Regulation of the vir genes of Agrobac-terium tumefaciens plasmid pTiC58. J Bacteriol 169:5101–5112.

Sadziene, A., Thomas, D.D., Bundoc, V.G., Holt, S.C., and

Barbour, A.G. (1991) A flagella-less mutant of Borreliaburgdorferi. J Clin Invest 88: 82–92.

Sambrook, J., Fritsch, E.F., and Maniatis, T. (1989) Molecu-lar Cloning: A Laboratory Manual. Cold Spring Harbor,New York: Cold Spring Harbor Laboratory Press.

Sanger, F., Nicklen, S., and Coulson, A.R. (1977) DNAsequencing with chain-terminating inhibitors. Proc NatlAcad Sci USA 74: 5463–5467.

Schroth, M.N., and Snyder, W.C. (1961) Effect of host exu-dates on chlamydospore germination of the bean root rotfungus, Fusarium solani f. phaseoli. Phytopathology 51:389–393.

Shaw, C.H., Loake, G.J., Brown, A.P., Garrett, G.S., Deakin,W., Alton, G., Hall, M., Jones, S.A., O’Leary, M., and Prim-vesi, L. (1991) Isolation and characterization of behav-ioural mutants and genes of Agrobacterium tumefaciens.J Gen Microbiol 137: 1939–1953.

Shirasu, K., and Kado, C.I. (1993) Membrane location of theTi plasmid VirB proteins involved in the biosynthesis of apilin-like conjugative structure on Agrobacterium tume-faciens. FEMS Microbiol Lett 111: 287–294.

Simon, R., Priefer, U., and Puhler, A. (1983) A broad hostrange mobilization system for in vivo genetic engineering:transposon mutagenesis in gram negative bacteria. Bio-technology 1: 784–791.

Smith, V.A., and Hindley, J. (1978) Effect of agrocin-84 onattachment of Agrobacterium tumefaciens to culturedtobacco cells. Nature 276: 498–500.

Tait, R.C., Close., T.J., Lundquist, R.C., Hagiya, M., Rodri-guez, R.L., and Kado, C.I. (1983) Construction and charac-terization of a versatile broad host range DNA cloningsystem for Gram-negative bacteria. Biotechnology 1: 269–275.

Tamura, Y., Kijima-Tanaka, M., Aoki, A., Ogikubo, Y., andTakahashi, T. (1995) Reversible expression of motilityand flagella in Clostridium chauvoei and their relationshipto virulence. Microbiology 141: 605–610.

Tanaka, S. (1985) Cytological studies on Agrobacteriumtumefaciens. 2. Flagella and their penetration canals.Yamaguchi Med Sci 34: 155–164.

Vande Broek, A., and Vanderleyden, J. (1995) The role ofbacterial motility, chemotaxis, and attachment in bacteria–plant interactions. Mol Plant–Microb Interact 8: 800–810.

Wassenaar, T.M., Bleumink-Pluym, N.M.C., and van de Zeijst,B.A.M. (1991) Inactivation of Camplylobacter jejuni flagel-lin genes by homologous recombination demonstratesthat flaA but not flaB is required for invasion. EMBO J10: 2055–2061.

Whitchurch, C.B., and Mattick, J.S. (1994) Characterizationof a gene, pilU, required for twitching motility but not phagesensitivity in Pseudomonas aeruginosa. Mol Microbiol 13:1079–1091.

Wu, J., Benson, A.K., and Newton, A. (1995) Global regula-tion of a s54-dependent flagellar gene family in Caulobactercrescentus by the transcriptional activator FlbD. J Bacteriol177: 3241–3250.

Zhang, M.Y., Lovgren, A., Low, M.G., and Landen, R. (1993)Characterization of an avirulent pleiotropic mutant of theinsect pathogen Bacillus thuringiensis reduced expressionofflagellinandphospholipases. Infect Immun61: 4947–4954.

590 O. Chesnokova et al.